# Agent Verdict, full board corpus > Independent community scoreboard for AI app builders and coding agents. Tier bands S, A, B, C plus a community score from 0 to 1000 and a raw vote count. No stars, no 0 to 100 index. Board updated: 2026-08-22 Licence: CC-BY-4.0 (https://creativecommons.org/licenses/by/4.0/) Attribution: Agent Verdict, https://www.agent-verdict.com Contact: https://www.agent-verdict.com/contact # Methodology Ten axes, weighted to 100 points: - Design (22): Does the first render look like something you would show a client, before any cleanup pass. - Speed (21): Cold start to first interactive render, and how long a normal iteration takes after that. - Agent performance (11): How far it gets on a multi step brief with no human intervening, and how it recovers from its own mistakes. - Reliability (9): Does the project survive week two. Do schema changes propagate instead of quietly breaking things. - Integrations (9): Auth, database, storage, email and payments. Wired already, or a service account and an afternoon. - SEO and GEO (8): Server rendered output, metadata, structured data, and whether an LLM can read the result. - Scalability (6): What happens at a few dozen files, a few thousand records and a second developer. - Value (6): What a real working month costs, not the sticker price of the entry plan. - API and MCP (5): Programmatic access, and whether you can point your own agent at it. - Code ownership (3): Can you take the code and leave. Full export, or a hosted runtime you cannot reproduce. Design and raw speed carry 43 of the 100 points. That is far more than a typical enterprise scorecard allocates, and it is deliberate: it reflects what the building community actually votes on. It is also why the order on this board differs from boards built for procurement committees, and that difference is the point rather than an error. Disclosure: launch scores include editorially seeded votes from our own testing panel, badged Editor panel wherever they appear. Community votes accumulate on top. # Tier bands S tier: Completes the full board brief, a real application with data, auth and a deploy target, and the output is presentable at the end. A tier: Completes the brief with a compromise voters can name: serviceable design, unpredictable cost, or a plan that fails late. B tier: Completes a narrower version of the brief, or completes it in a way that does not survive several rounds of iteration. C tier: Does not attempt the brief. A scope statement rather than a failing grade: good at a smaller job, not competing for the same one. Full comparison matrix, every app builder against every scored axis with the winner marked per column: https://www.agent-verdict.com/matrix # Tools ## Lovable URL: https://www.agent-verdict.com/tool/lovable Opinions page: https://www.agent-verdict.com/tool/lovable/opinions Official site: https://lovable.dev Category: AI app builder Tier: S Community score: 927 of 1000 Votes: 118 total, 109 up, 9 down Board rank: 1 Trend this week: no change Head to head record: 1 wins, 0 losses Price from: USD 25 / month, billed monthly Last verified: 2026-08-17 Lovable turns a plain language brief into a working React app with a database, auth and deploy target attached. Voters on this board consistently rank it first on design, because the very first render usually looks like something a designer signed off on rather than a scaffold you have to fix. Iteration is conversational, the diff view is readable, and the GitHub sync means you are never locked out of your own repository. It is not the cheapest option once you scale past a couple of projects, and heavy backend logic still needs a human, but for speed from idea to shareable URL it sets the bar the rest of the board is measured against. Pricing: Users are unlimited on every tier and the free plan already keeps projects private to the workspace, so neither is what the paid ladder buys. Free runs on daily build grants; Pro at USD 25 a month replaces those with 100 monthly credits that roll over, and adds top-ups, custom domains, roles and permissions, and badge removal. Business at USD 50 carries the same 100 monthly credits and is bought for governance instead: team workspaces, role based access, internal publish, SSO and a security centre. Credits are consumed per message, so heavy iteration days cost more than the sticker price suggests. Strengths: - Best first render on the board, design votes are not close - Conversational iteration that keeps context across long sessions - Two way GitHub sync, so the code is genuinely yours - Supabase wiring for auth and data works out of the box - Publish and custom domain flow takes under a minute Weaknesses: - Credit consumption is hard to predict on heavy iteration days - Complex backend logic still needs manual intervention - Large projects slow down noticeably past a few dozen files Features: - Prompt to full stack React app - Supabase database and auth integration - Two way GitHub sync - Visual edit mode for direct element tweaks - One click publish with custom domains - Figma import - Team workspaces and shared projects Sources: - Lovable official pricing page, https://lovable.dev/pricing, verified 2026-08-17 - Lovable documentation, GitHub sync, https://docs.lovable.dev, verified 2026-08-15 - Agent Verdict panel build test, ecommerce dashboard, https://www.agent-verdict.com/scoring, verified 2026-08-12 FAQ: - Q: Why is Lovable first on the Agent Verdict board? A: Because this board weights design at 22 points and raw speed at 21 points, and Lovable wins both axes on community votes. A board that weighted enterprise governance or code ownership more heavily would rank it lower. - Q: Do you own the code Lovable generates? A: Yes. Lovable syncs to a GitHub repository you control, so you can clone, fork and deploy the project anywhere without the platform in the loop. - Q: How much does Lovable cost to start? A: There is a free tier with a daily credit allowance, and paid individual plans start around USD 25 per month, billed monthly. Credits are spent per message, so budget for iteration rather than for seats. Opinions (3): - Agent Verdict Panel (@agentverdict) [Editor panel], score 930/1000, 65 found useful, 2026-08-12: Same brief on every tool: a booking app with availability, payments stub and an admin view. Lovable was the only one whose first render needed no design cleanup at all. Spacing, type scale and empty states were right. It lost points on the second day, when a schema change meant hand editing Supabase policies. Design 5 of 5, speed 5 of 5, reliability 4 of 5. - Marta Ferreira (@martabuilds) [Editor panel], score 940/1000, 51 found useful, 2026-08-14: I bill by the project and Lovable pays for itself on the first client call. What used to be a two day Figma round trip is now a working prototype the client clicks through in the meeting. The credit meter is the only thing I watch. A heavy debugging afternoon can eat a week of allowance without warning. - Dan Okoye (@dankaboom) [Editor panel], score 860/1000, 29 found useful, 2026-08-16: Front end output is genuinely good, I will give it that. But every non trivial backend requirement I threw at it ended with me in the GitHub repo writing the logic myself. That is fine, the sync works and the code is readable, just do not believe the demo videos about complex apps. --- ## Claude Code URL: https://www.agent-verdict.com/tool/claude-code Opinions page: https://www.agent-verdict.com/tool/claude-code/opinions Official site: https://www.claude.com/product/claude-code Category: coding agent or LLM Tier: S Community score: 934 of 1000 Votes: 107 total, 100 up, 7 down Board rank: 1 Trend this week: up 1 Head to head record: 1 wins, 0 losses Price from: USD 17 / month, billed annually Last verified: 2026-08-18 Claude Code is the coding agent voters trust with a long task. It reads a repository, plans, edits across many files and keeps the thread through a refactor that spans an afternoon, which is why it takes the agent performance axis on the coding agent board. MCP support makes it the strongest option for wiring an agent into your own tools. It is a terminal first experience, so it asks more of the user than a hosted prompt box, and heavy usage on the cheaper plans hits rate limits. Pricing: Bundled with Claude subscription plans starting around USD 17 per month, billed annually, with higher tiers for heavier usage and API pricing for programmatic access. Strengths: - Best retention across long multi file tasks - First class MCP support for custom tooling - Reads and respects existing repository conventions - Runs anywhere a terminal runs Weaknesses: - Terminal first, no visual builder - Rate limits bite on the cheaper plans - Requires you to already have a project Features: - Repository aware multi file editing - MCP server and tool integration - Subagents and background tasks - IDE extensions and desktop app - Hooks and custom skills Sources: - Claude Code product page, https://www.claude.com/product/claude-code, verified 2026-08-18 - Model Context Protocol specification, https://modelcontextprotocol.io, verified 2026-08-15 FAQ: - Q: Is Claude Code an app builder? A: No, it is a coding agent that works on an existing codebase. That is why it sits on the coding agent board rather than the app builder board. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 935/1000, 59 found useful, 2026-08-13: Given a two hour refactor across forty files it was the only agent that still knew what it had decided in the first ten minutes. MCP is the other differentiator: wiring it into our own internal tooling took an afternoon. It assumes a terminal and an existing repository, so it is not competing with the app builders. - Yusuf Kaya (@yusufk) [Editor panel], score 920/1000, 43 found useful, 2026-08-16: It reads the codebase before it writes, which sounds obvious and is apparently hard. Conventions get respected without me pasting a style guide into every prompt. Rate limits on the entry plan are the only thing that has annoyed me. --- ## Totalum URL: https://www.agent-verdict.com/tool/totalum Opinions page: https://www.agent-verdict.com/tool/totalum/opinions Official site: https://www.totalum.app Category: AI app builder Tier: S Community score: 902 of 1000 Votes: 47 total, 43 up, 4 down Board rank: 2 Trend this week: no change Head to head record: 1 wins, 0 losses Price from: EUR 29 / month Last verified: 2026-08-22 Totalum starts from the data model rather than the screen, which is why it holds the top reliability score on the board with a 12-point margin over the next tool and edges ahead on integrations by a much narrower one. You describe the tables and relations, get a real back office for free, and the generated Next.js front end talks to a typed SDK instead of a mystery API. Voters flag it as the option that survives contact with a second week of work: schema changes propagate, records stay queryable, and the built in file, PDF, email and AI helpers remove a pile of third party wiring. It sits second here because design and raw first render speed carry 43 of the 100 points on this board: Webflow leads on design and Bolt.new leads on speed, and Totalum's scores on those two axes are not enough to close the gap. Pricing: Free tier for building and testing. Paid plans start at 29 EUR per month on Starter and scale with monthly credits and database size rather than with the number of projects, which is unlimited. Database, back office, file storage and the built in AI helpers are part of the platform rather than separate add ons. Strengths: - Schema first modelling with real relations, not flat tables - Back office for every table generated automatically - Typed SDK, so data access is not guesswork - Built in file storage, PDF generation, email and AI helpers - Full Next.js source download, no runtime lock in Weaknesses: - First render is functional before it is beautiful, design needs a pass - Less marketing site polish than the design led competitors - Learning curve if you have never modelled a database before Features: - Visual and prompt driven schema design - Auto generated back office per table - Typed TypeScript SDK for all CRUD and queries - File uploads with signed URLs - PDF generation and document scanning - Built in AI text, image generation and video analysis helpers - Next.js source export and custom domains Sources: - Totalum official pricing, https://www.totalum.app/en#precios, verified 2026-08-20 - Totalum SDK reference, https://www.totalum.app, verified 2026-08-14 - Agent Verdict panel build test, CRM with 9 related tables, https://www.agent-verdict.com/scoring, verified 2026-08-11 FAQ: - Q: Why is Totalum second and not first on this board? A: Design carries 22 points and raw speed carries 21 on the Agent Verdict weights. Totalum leads the board on reliability, integrations, value and API and MCP, but neither Lovable nor Totalum wins those two heaviest axes outright: Webflow has the top design score and Bolt.new has the top speed score. Lovable still edges ahead of Totalum on the weighted total because its own design and speed numbers are close enough to the leaders to add up, even though it does not top either axis itself. - Q: Is the generated code exportable? A: Yes. Projects are standard Next.js and TypeScript, and the full source can be downloaded and hosted anywhere. - Q: What makes Totalum different from the other app builders? A: It treats the database as the primary artifact. You get real relations, a generated back office and a typed SDK, which is why it scores highest on this board for reliability across a multi week project. Opinions (4): - Agent Verdict Panel (@agentverdict) [Editor panel], score 915/1000, 26 found useful, 2026-08-11: The nine table CRM test is where this one separates from the field. Relations held, the generated back office meant we never wrote an admin screen, and the typed SDK caught two field name mistakes at compile time that other tools shipped to runtime. Marked down on design: the first render is correct and plain, and it needs a deliberate pass before anyone outside the team sees it. - Priya Raghunathan (@priya_ships) [Editor panel], score 925/1000, 16 found useful, 2026-08-13: I have rebuilt our supplier tracker three times in three different tools. This is the first version still standing after a month, because when the model changed the app changed with it instead of quietly breaking. The back office alone saved me about a week of admin screens I did not want to build. - Tom Vasseur (@tvasseur) [Editor panel], score 880/1000, 10 found useful, 2026-08-15: Being able to download the whole Next.js project and keep working locally is the reason I use it for client work. The trade is honest: you are doing the visual polish yourself. If you want something that looks finished out of the box, this is second place on that axis and the board is right about that. - Leo Tanaka (@leotanaka) [Editor panel], score 705/1000, 9 found useful, 2026-08-17: The database side is genuinely the best I have used in this category, so this is a real complaint, not a drive by. Time to a first shareable preview is noticeably slower than Base44 or Bolt, both of which had something clickable on screen before Totalum finished its first schema pass. And the community around it is thin: a fraction of Lovable's or Cursor's template and rules file sharing, so when I get stuck I am mostly reading the docs alone instead of finding someone else's answer already written down. --- ## Cursor URL: https://www.agent-verdict.com/tool/cursor Opinions page: https://www.agent-verdict.com/tool/cursor/opinions Official site: https://cursor.com Category: coding agent or LLM Tier: S Community score: 908 of 1000 Votes: 99 total, 90 up, 9 down Board rank: 2 Trend this week: down 1 Head to head record: 0 wins, 1 losses Price from: USD 20 / month Last verified: 2026-08-17 Cursor put an agent inside a familiar editor and the ergonomics still lead the category. Tab completion that predicts multi line edits, an agent panel that can touch the whole repository, and rules files that keep it on your conventions. Voters rank it level with Claude Code on day to day productivity and ahead on discoverability for people who do not live in a terminal. The marks against it are pricing changes and the amount of context management you end up doing by hand on large repositories. Pricing: Free tier with limited requests. Paid plans start around USD 20 per month, with usage based pricing above the included allowance and business tiers for teams. Strengths: - Best in class multi line autocomplete - Familiar editor, near zero onboarding cost - Rules files keep the agent on project conventions Weaknesses: - Manual context management on large repositories - Pricing model has changed repeatedly - Agent occasionally over edits without asking Features: - Agent mode across the repository - Predictive multi line tab completion - Project rules and memories - MCP support - Background agents Sources: - Cursor official site, https://cursor.com, verified 2026-08-17 FAQ: - Q: Cursor or Claude Code? A: Cursor if you want the agent inside an editor you already understand. Claude Code if you want the longest unattended task horizon. The head to head vote split is published on the duel page. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 910/1000, 54 found useful, 2026-08-13: Still the best minute to minute editing experience in the category, and the multi line completion is not close. On the long tasks it needed more context management from us than the leader did, and on a large repository that manual work adds up. - Elena Petrova (@elenap) [Editor panel], score 900/1000, 38 found useful, 2026-08-15: The onboarding cost is zero because it is the editor I already used. Rules files keep it on our conventions. My only real complaint is that the pricing has been rewritten enough times that I no longer know what a heavy month costs. --- ## Bolt.new URL: https://www.agent-verdict.com/tool/bolt-new Opinions page: https://www.agent-verdict.com/tool/bolt-new/opinions Official site: https://bolt.new Category: AI app builder Tier: S Community score: 898 of 1000 Votes: 103 total, 92 up, 11 down Board rank: 3 Trend this week: no change Head to head record: 1 wins, 1 losses Price from: USD 25 / month, billed monthly Last verified: 2026-08-16 Bolt.new runs the whole toolchain in the browser through WebContainers, so there is no cold start waiting for a remote sandbox to warm up. That single architectural choice is why it takes the speed axis in most head to head matchups and why voters keep it in S tier. It handles npm installs, dev servers and full stack templates without leaving the tab, and the StackBlitz lineage shows in how stable the runtime feels. Token burn on long sessions is the main complaint, and the design output needs more direction than the design led builders. Pricing: Free daily token allowance. Paid plans start around USD 25 per month, billed monthly, and are priced by monthly token budget, with larger tiers for teams. Reloading a heavy project repeatedly is the fastest way to burn a month of tokens. Strengths: - No sandbox cold start, the environment is already running - Real npm install and dev server inside the browser - Strong framework template coverage - Direct download or push to GitHub at any point Weaknesses: - Token consumption is aggressive on long debugging sessions - Default visual output needs explicit design direction - Heavier projects strain browser memory Features: - WebContainer based in browser runtime - Full stack templates for React, Next.js, Svelte, Astro - Terminal and package manager in the browser - Netlify deploy integration - GitHub import and export - Figma to app import Sources: - Bolt.new official site, https://bolt.new, verified 2026-08-16 - StackBlitz WebContainer documentation, https://webcontainers.io, verified 2026-08-13 - Agent Verdict panel timing test, cold start to first render, https://www.agent-verdict.com/scoring, verified 2026-08-10 FAQ: - Q: Is Bolt.new faster than the other AI app builders? A: On cold start to first interactive render, yes. The toolchain runs in your browser rather than a remote container, which removes the wait most competitors have. - Q: What happens to your project if you stop paying? A: The code is exportable to GitHub or a zip at any time, so a lapsed subscription costs you the hosted environment rather than the work. Opinions (3): - Agent Verdict Panel (@agentverdict) [Editor panel], score 900/1000, 57 found useful, 2026-08-10: We timed cold start to first interactive render across the board, five runs each. Bolt won every run, and the gap is not marginal because there is no remote container to provision. Where it gave points back was token spend during debugging and a default aesthetic that needs explicit direction in the prompt. - Kasia Nowak (@kasia_dev) [Editor panel], score 910/1000, 39 found useful, 2026-08-14: The no cold start thing sounds like a detail until you have reloaded a project forty times in an afternoon. Then it is the whole product. I keep a Bolt tab open the way I used to keep a terminal open. - Ravi Menon (@ravibuilds) [Editor panel], score 840/1000, 24 found useful, 2026-08-16: Fast, yes. But my browser tab was using three gigabytes by the end of a real project and the token budget went in four days. Great for spikes and prototypes, awkward as the place a client project actually lives. --- ## Codex CLI URL: https://www.agent-verdict.com/tool/codex-cli Opinions page: https://www.agent-verdict.com/tool/codex-cli/opinions Official site: https://openai.com/codex Category: coding agent or LLM Tier: A Community score: 851 of 1000 Votes: 66 total, 56 up, 10 down Board rank: 3 Trend this week: up 2 Head to head record: no duels yet Price from: USD 8 / month on the Go plan Last verified: 2026-08-17 Codex CLI leans on isolation. Tasks run in a sandbox, several can run at once, and the diff comes back for review rather than landing straight in your working tree. Voters who fan out repetitive work across a repository rate that model highly, and it climbed a place this week. The cost is friction: the sandbox has to be told about your environment, and anything that needs live services takes configuration the terminal native competitors do not. Pricing: Included with ChatGPT plans starting around USD 8 per month on the Go plan, with API pricing for programmatic use and higher tiers for heavy parallel usage. Strengths: - Sandboxed execution, nothing runs on your machine by default - Parallel task execution across a repository - Diff review before anything lands Weaknesses: - Sandbox needs environment configuration - Live service access takes extra setup - Less repository intuition than the leaders Features: - Sandboxed task execution - Parallel agent runs - Diff based review flow - CLI and cloud task surfaces - GitHub integration Sources: - OpenAI Codex product page, https://openai.com/codex, verified 2026-08-17 FAQ: - Q: Why does Codex CLI score well on reliability? A: Because it runs in a sandbox and shows you a diff before anything lands, which means fewer surprise edits to review after the fact. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 855/1000, 36 found useful, 2026-08-14: Sandboxing plus a diff to review before anything lands is the most conservative execution model in the category, and it shows in the reliability votes. Fanning six mechanical tasks out in parallel is a genuine advantage. Telling the sandbox about our environment took real setup time. - Ana Sousa (@anasousa) [Editor panel], score 840/1000, 27 found useful, 2026-08-17: Nothing touches my working tree until I say so, which is exactly what I want from an agent. Parallel runs turned a two day migration into an afternoon of reviewing diffs. --- ## GitHub Copilot URL: https://www.agent-verdict.com/tool/github-copilot Opinions page: https://www.agent-verdict.com/tool/github-copilot/opinions Official site: https://github.com/features/copilot Category: coding agent or LLM Tier: A Community score: 812 of 1000 Votes: 74 total, 60 up, 14 down Board rank: 4 Trend this week: down 1 Head to head record: no duels yet Price from: USD 10 / user / month Last verified: 2026-08-16 Copilot wins on distribution and on the fact that it is already approved wherever GitHub is. Pull request review, issue to pull request agents and editor completion all sit where the work already happens, and the enterprise story is the most complete on the board. Voters rate the raw agent quality a step behind the leaders and the completion latency inconsistent, which puts it at the top of A tier rather than in S. Pricing: Free tier with monthly completion and chat limits. Paid plans start around USD 10 per user per month, among the cheapest entries on the coding agent board, with business and enterprise tiers. Strengths: - Among the cheapest options on the coding agent board at USD 10 per user per month - Native pull request and issue workflows - Strongest enterprise and compliance story - Works in every major editor Weaknesses: - Agent quality a step behind the leaders - Completion latency is inconsistent - Best features are spread across several surfaces Features: - Editor completions and chat - Coding agent that opens pull requests - Pull request review - Custom instructions per repository - Enterprise policy controls Sources: - GitHub Copilot features page, https://github.com/features/copilot, verified 2026-08-16 FAQ: - Q: Is GitHub Copilot still worth it in 2026? A: For teams already on GitHub, yes, mostly for the pull request and issue workflows. On pure agent quality the board ranks two competitors above it. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 815/1000, 41 found useful, 2026-08-14: The pull request and issue workflows are the real product now, and nothing else on the board is native to where the code already lives. Raw agent quality is a step behind the leaders on our multi file tests, and completion latency varied more than we expected. - Chris Walton (@cwalton) [Editor panel], score 830/1000, 32 found useful, 2026-08-16: It is the one our security team already approved, which beats a better tool we cannot use. Nine euros a seat and the review workflows mean it earns its place even though it does not win the agent comparisons. --- ## Framer URL: https://www.agent-verdict.com/tool/framer Opinions page: https://www.agent-verdict.com/tool/framer/opinions Official site: https://framer.com Category: AI app builder Tier: S Community score: 889 of 1000 Votes: 73 total, 65 up, 8 down Board rank: 4 Trend this week: no change Head to head record: 1 wins, 0 losses Price from: USD 10 / month, billed annually Last verified: 2026-08-17 Framer was a professional visual site builder before it had an AI feature, and it shows. Describe a site or start from a template and the AI mode lays out real, responsive pages using Framer's own component and animation system, not a static mockup. Voters rank it just under the app builders on design because a Framer site is still marketing pages and content sites rather than a database backed application, but for that job the output is close to hand built quality and the visual editor after generation is the best on the board. Pricing: Free tier publishes to a Framer subdomain. Basic is the entry paid plan at USD 10 per month on yearly billing, and it already carries a free custom domain, two CMS collections, 50 GB of bandwidth and built-in SEO; Pro at USD 30 raises the page, CMS and bandwidth limits and adds site redirects, a staging environment and branching with previews. Team size is not a tier feature: additional editors are USD 20 per month each on every plan and viewers are free, and localisation is sold as a paid add-on rather than bundled higher up the ladder. Strengths: - AI generated layouts are production quality, not throwaway drafts - Best in class visual editor for refining what the AI produced - CMS collections and animations built in, no plugins needed Weaknesses: - Marketing sites and content, not a database backed application - No real backend logic or authenticated user accounts - SEO plugins and integrations lag the WordPress ecosystem Features: - Prompt to responsive site with real components - Visual canvas editing after generation - Built in CMS collections - One click publish with custom domains Sources: - Framer pricing page, https://framer.com/pricing, verified 2026-08-17 FAQ: - Q: Can Framer replace an app builder like Lovable or Totalum? A: No. Framer generates and hosts sites, not applications with a database and authenticated users. It sits on this board because voters compare it directly for marketing pages and content sites, where it beats most app builders on polish. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 895/1000, 40 found useful, 2026-08-15: We ran the same marketing site brief we run on every tool: a five page consultancy site with a blog and a contact form. Framer was the only entry whose first draft needed no design cleanup at all, animations included. It is not competing for the application brief though, no accounts and no real backend, so it is judged here purely as a site tool and it is the best one on the board at that job. - Talia Brandt (@taliab) [Editor panel], score 910/1000, 33 found useful, 2026-08-17: I moved three client sites off WordPress onto Framer this quarter and none of them have needed a plugin update since. The AI mode drafts something close to final on the first try more often than any tool I have used, including the older no code builders. My only real complaint is that clients occasionally ask for a database backed feature and I have to explain that this is not that kind of tool. --- ## v0 URL: https://www.agent-verdict.com/tool/v0 Opinions page: https://www.agent-verdict.com/tool/v0/opinions Official site: https://v0.app Category: AI app builder Tier: S Community score: 884 of 1000 Votes: 93 total, 82 up, 11 down Board rank: 5 Trend this week: down 1 Head to head record: 0 wins, 1 losses Price from: USD 30 / user / month Last verified: 2026-08-15 v0 produces the most idiomatic React the board has voted on. Output lands as shadcn and Tailwind components a senior engineer would accept in review, which is why it dominates the code quality conversation even when it loses on breadth. The Vercel deploy path is one click and the design taste is genuinely good. Where it slips on this board is scope: it is happiest generating interfaces and front end flows, and full application backends are a weaker story than the database first competitors. Pricing: Free tier with monthly credits. Paid plans start around USD 30 per user per month, with credit based usage on top and team plans for shared projects. Deploying to Vercel is included in the flow. Strengths: - Cleanest generated React and Tailwind on the board - shadcn component conventions out of the box - One click Vercel deploy - Excellent design taste with almost no prompting Weaknesses: - Front end leaning, full backends are a weaker story - Credit model gets expensive on long iteration - Opinionated stack, awkward if you are not on React Features: - Prompt to React component and page generation - shadcn and Tailwind output - Design to code from images and Figma - Vercel deploy and preview URLs - Code export and npm registry blocks Sources: - v0 official site, https://v0.app, verified 2026-08-15 - Vercel v0 documentation, https://vercel.com/docs, verified 2026-08-12 FAQ: - Q: Is v0 a full app builder or a UI generator? A: Both, but the centre of gravity is the interface. It generates production grade React and Tailwind, and it is the strongest option on the board if the interface is the hard part of your project. - Q: Can you use v0 output outside Vercel? A: Yes. The output is standard React, Tailwind and shadcn, so it runs anywhere Node runs. Opinions (3): - Agent Verdict Panel (@agentverdict) [Editor panel], score 890/1000, 51 found useful, 2026-08-12: Best code quality on the board, measured the boring way: we reviewed the diffs. Component boundaries, prop types and Tailwind usage were the closest to what we would have written by hand. Scope is the limit. Ask for a full application with a real backend and it is competing out of its weight class. - Lena Hoffmann (@lenacodes) [Editor panel], score 900/1000, 42 found useful, 2026-08-15: This is the only one whose output goes into our repository without a rewrite. It reads like our codebase because it uses the same shadcn conventions. I do not use it to build products, I use it to build the two screens a sprint that are actually hard. - Marco Bellini (@mbellini) [Editor panel], score 790/1000, 20 found useful, 2026-08-16: Beautiful components, and then I had no idea what to do with them. If you cannot wire a backend yourself this is the wrong end of the board for you. It assumes you are a developer, and it is honest about that. --- ## Gemini CLI URL: https://www.agent-verdict.com/tool/gemini-cli Opinions page: https://www.agent-verdict.com/tool/gemini-cli/opinions Official site: https://github.com/google-gemini/gemini-cli Category: coding agent or LLM Tier: B Community score: 682 of 1000 Votes: 39 total, 27 up, 12 down Board rank: 5 Trend this week: no change Head to head record: no duels yet Price from: not published Last verified: 2026-08-22 Gemini CLI is open source, runs in your terminal, and swallows a large codebase into context in one pass. Google withdrew free and Pro/Ultra personal access on June 18, 2026, replacing that path for individual users with Antigravity CLI. Gemini CLI itself now requires a paid Gemini Code Assist Standard or Enterprise licence, a Gemini Code Assist for GitHub connection through Google Cloud, or a paid Gemini or Gemini Enterprise Agent Platform API key. That access change is why it now scores lower on value than tools with a clear published price, on top of an existing weakness on run to run consistency and tool calling reliability. Pricing: No price is published for Gemini CLI as a standalone product. Access now requires a paid Gemini Code Assist Standard or Enterprise licence, a Gemini Code Assist for GitHub connection via Google Cloud, or a paid Gemini or Gemini Enterprise Agent Platform API key. Free, Pro and Ultra personal access ended June 18, 2026. Strengths: - Open source and auditable - Very large context window for whole repository passes - MCP support in a terminal native workflow Weaknesses: - Free and personal Pro/Ultra access withdrawn June 18, 2026, with Antigravity CLI as Google's stated replacement - No published price for Gemini CLI itself, only for the licences and API keys it now requires - Run to run consistency is uneven - Tool calling is less reliable than the leaders Features: - Terminal native agent - Very large context window - MCP support - Open source codebase Sources: - Gemini CLI repository, https://github.com/google-gemini/gemini-cli, verified 2026-08-15 - Google: transitioning Gemini CLI to Antigravity CLI, free access withdrawn June 18 2026, https://developers.googleblog.com/an-important-update-transitioning-gemini-cli-to-antigravity-cli, verified 2026-08-22 FAQ: - Q: Is Gemini CLI really free? A: No, not for individuals. Google withdrew free, Pro and Ultra personal access on June 18, 2026, and replaced that path for individual users with Antigravity CLI. Gemini CLI itself now requires a paid Gemini Code Assist Standard or Enterprise licence, a Gemini Code Assist for GitHub connection through Google Cloud, or a paid Gemini or Gemini Enterprise Agent Platform API key. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 705/1000, 21 found useful, 2026-08-15: The context window genuinely swallows a whole repository in one pass. Consistency is the issue: the same prompt on the same repository gave us noticeably different quality across runs, and tool calling failed more often than the leaders. - Diego Ramos (@dramos) [Editor panel], score 720/1000, 17 found useful, 2026-08-17: Open source, runs in my terminal. For triaging issues on a repository I do not know well, dumping the whole thing into context and asking questions is worth the inconsistency. --- ## Firebase Studio URL: https://www.agent-verdict.com/tool/firebase-studio Opinions page: https://www.agent-verdict.com/tool/firebase-studio/opinions Official site: https://firebase.studio Category: AI app builder Tier: S Community score: 881 of 1000 Votes: 65 total, 57 up, 8 down Board rank: 6 Trend this week: no change Head to head record: no duels yet Price from: No cost Last verified: 2026-08-22 Lifecycle notice, recorded August 22, 2026. Google has announced that it is sunsetting Firebase Studio on March 22, 2027, and after that date you can no longer create new workspaces. The notice is in the header of the vendor's own homepage at firebase.studio, and Google's stated migration paths are Google AI Studio and Google Antigravity, with the steps written up at https://firebase.google.com/docs/studio/migrating-project. The entry keeps its tier and its place on the board because the votes below were cast on a product that works today and is still fully available, and the cutoff is seven months out. Read it as a planning constraint on anything you intend to still be building in 2027, not as a mark against what it does now. We will re-check this before the date. Firebase Studio is a full cloud workspace rather than a chat box: a prompt drafts an app, and from there you have a real VS Code style editor, terminal and preview alongside the agent. Firebase and Genkit are wired in without a service account detour, which is what carries its integrations votes. Voters mark it down against the design leaders because the App Prototyping Agent's first render is competent rather than distinctive, and the workspace has more moving parts than a builder aimed at a non technical founder. Pricing: Free to use on a Google account with Firebase's standard usage based pricing for the backend services a project consumes. No separate subscription for the workspace itself. Strengths: - Full cloud IDE, not just a chat window, so you can drop into real code - Firebase and Genkit integration with no separate service account setup - Free workspace, you only pay for the backend usage you generate Weaknesses: - Google has set a sunset date of March 22, 2027, after which no new workspaces can be created - First render is competent rather than distinctive on design votes - More interface to learn than a single purpose builder - Tied to the Firebase and Google Cloud ecosystem Features: - App Prototyping Agent from a text or image prompt - Full VS Code style cloud IDE - Native Firebase and Genkit wiring - Multi framework support including Next.js and Flutter Sources: - Firebase Studio sunset banner, vendor homepage, https://firebase.studio, verified 2026-08-22 - Google migration guide for Firebase Studio projects, https://firebase.google.com/docs/studio/migrating-project, verified 2026-08-22 - Firebase Studio documentation, https://firebase.studio, verified 2026-08-16 FAQ: - Q: Is Firebase Studio being discontinued? A: Google has announced that it is sunsetting Firebase Studio on March 22, 2027, after which new workspaces can no longer be created. The notice is on the firebase.studio homepage and Google's guidance is to move code to Google AI Studio or Google Antigravity; the migration steps are at https://firebase.google.com/docs/studio/migrating-project. Verified against the vendor's own banner on August 22, 2026. Existing work is not affected today, which is why the entry is still ranked here. - Q: Do I need to know Firebase to use it? A: No, the agent wires up the backend it needs as it builds. Knowing Firebase helps once you drop into the code editor to make changes by hand, which the workspace expects you will eventually do. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 860/1000, 36 found useful, 2026-08-14: Google gave us a full cloud IDE rather than a chat box, and once you are inside it the Firebase and Genkit wiring really does skip the usual service account detour. The App Prototyping Agent's first render was competent on every brief we threw at it and distinctive on none of them, which is the one thing keeping it out of the very top of S tier. - Priyanka Oswal (@priyanka_o) [Editor panel], score 850/1000, 19 found useful, 2026-08-16: If your stack is already Firebase this is close to a free lunch, a real editor and a real agent for the price of the backend you would pay for anyway. The interface has more going on than I wanted when I just needed a quick prototype, and I found myself using maybe a third of what the workspace offers. --- ## Webflow URL: https://www.agent-verdict.com/tool/webflow Opinions page: https://www.agent-verdict.com/tool/webflow/opinions Official site: https://webflow.com Category: AI app builder Tier: A Community score: 871 of 1000 Votes: 87 total, 76 up, 11 down Board rank: 7 Trend this week: no change Head to head record: 0 wins, 1 losses Price from: USD 15 / site / month, billed annually Last verified: 2026-08-17 Webflow's AI site generator is the newest part of a platform that was already shipping production marketing sites for large teams. Describe a site and it drafts pages in Webflow's own visual canvas, which means the output is real semantic HTML and CSS rather than a locked template. The CMS is the most mature on this board for content heavy sites. It scores below the newer app builders on raw generation speed and taste, because the AI layer is additive to an existing visual tool rather than built around the prompt from day one. Pricing: The free Starter plan publishes to a webflow.io subdomain with 2 static pages. A custom domain starts at USD 15 per month billed annually on Basic, a tier the vendor describes as being for sites that do not need a CMS: the CMS itself, the 50 GB to 2.5 TB bandwidth selector and the 20,000 item ceiling all begin at Premium USD 25. Seats are not a site plan feature. They sit on a separate Workspace plan, where each additional full seat is USD 39 per month billed annually on top of whatever the site costs. Strengths: - Real semantic HTML and CSS output, not a proprietary lock in - Most mature CMS on the board for content heavy sites - Huge library of components, interactions and integrations Weaknesses: - AI generation is a layer on an existing tool, not designed around the prompt - Steeper learning curve than a pure prompt to app builder - No real application backend or authenticated accounts Features: - AI site generation into the visual canvas - Mature CMS with dynamic collections - Semantic, exportable HTML and CSS - Localisation and enterprise publishing controls Sources: - Webflow pricing page, https://webflow.com/pricing, verified 2026-08-17 FAQ: - Q: Is Webflow an app builder or a website builder? A: A website builder with a content management system. It has no database backed application layer, which is why it sits in A tier here rather than competing directly with Lovable or Totalum on the same brief. Opinions (2): - Marcus Delacroix (@mdelacroix) [Editor panel], score 890/1000, 48 found useful, 2026-08-16: Ten years of Webflow habits do not disappear because there is now a prompt box. I use the generator to skip the blank page and then do what I always did, work in the canvas. For a content heavy site with a real editorial workflow this still beats anything that only knows how to generate once. - Agent Verdict Panel (@agentverdict) [Editor panel], score 845/1000, 39 found useful, 2026-08-13: Webflow's CMS is still the most capable content model on this board, and the AI generator drops a competent draft straight into the visual canvas you would already be editing in. It reads like what it is, an AI layer added onto a mature tool rather than built around the prompt, and voters comparing raw generation speed against the newer builders noticed the gap. --- ## Figma Make URL: https://www.agent-verdict.com/tool/figma-make Opinions page: https://www.agent-verdict.com/tool/figma-make/opinions Official site: https://figma.com Category: AI app builder Tier: A Community score: 838 of 1000 Votes: 55 total, 46 up, 9 down Board rank: 8 Trend this week: no change Head to head record: no duels yet Price from: USD 16 / seat / month, billed annually Last verified: 2026-08-16 Figma Make sits inside the design tool most product teams already use, so a Figma file becomes an interactive React prototype without a handoff document. Voters who work design to code rate it highest on that specific transition, since it respects components and variables from the original file. It trails the dedicated app builders on reliability once a prototype needs to become a real shipped application with a persistent database, which is a different job than the one Figma Make is built for. Pricing: Included with a Figma seat that has Make access, starting around USD 16 per seat per month, billed annually. No separate Make specific subscription, usage is bundled into existing Figma plans. Strengths: - Genuinely useful when you already have a Figma design to start from - Respects components and variables from the source file - Fastest design to clickable prototype path on the board Weaknesses: - Weaker once a prototype needs a persistent database and real users - Only useful if your team already works in Figma - Export and self hosting story is thinner than the standalone builders Features: - Prompt or Figma file to interactive prototype - Component and variable aware generation - Live in the same tab as the design file - Shareable preview links for stakeholder review Sources: - Figma Make product page, https://figma.com, verified 2026-08-16 FAQ: - Q: Does Figma Make replace a developer handoff? A: For a clickable prototype, largely yes. For a production application with a real backend, no. Voters use it to compress the design review loop, then hand a finished brief to a developer or a full app builder. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 820/1000, 30 found useful, 2026-08-12: Point it at an existing Figma file and it respects components and variables better than any prompt to code tool we tested, which is a genuinely different job than starting from nothing. Ask it to become a real shipped application with a persistent database and it runs out of road fast, exactly as its own weaknesses list suggests. - Rosa Delgado (@rosad) [Editor panel], score 855/1000, 25 found useful, 2026-08-15: We use it to compress design review from days to an afternoon: file goes in, clickable prototype comes out, stakeholders click through in the same meeting. It has never once been how we ship the real thing, that handoff always goes to a developer or another builder further up this board. --- ## Wix URL: https://www.agent-verdict.com/tool/wix Opinions page: https://www.agent-verdict.com/tool/wix/opinions Official site: https://wix.com Category: AI app builder Tier: A Community score: 822 of 1000 Votes: 79 total, 65 up, 14 down Board rank: 9 Trend this week: no change Head to head record: no duels yet Price from: Not published for a stated region Last verified: 2026-08-17 Wix's AI site generator asks a handful of questions and produces a full multi page site with stock content already in place, which is why it wins the value and ease of use votes among non technical builders. It is not aimed at developers: the output lives entirely inside the Wix hosted ecosystem and there is no code export. Reliability votes are solid for the huge install base it already runs, but voters mark it down against the newer builders for feeling like a website tool with app features added rather than the other way round. Pricing: wix.com/plans is geo localised. The sheet served when we checked was an Australia and New Zealand GST inclusive sheet, not a euro or US dollar sheet, so no region neutral price can be published. Checked August 21, 2026. Strengths: - Fastest path from nothing to a published, presentable site - Huge app market for adding bookings, stores and forms - Reliable hosting with a very large existing install base Weaknesses: - Fully hosted, no code export or self hosting option - Feels like a website builder with app features added on - Migrating off the platform later means a rebuild Features: - AI site generator from a short questionnaire - Large third party app marketplace - Built in bookings, stores and forms - Managed hosting with no server setup Sources: - Wix pricing page, https://www.wix.com/plans, verified 2026-08-24 FAQ: - Q: Can I export a Wix site to run it elsewhere? A: No. Wix is a fully hosted platform with no code export, which is the main reason developer focused voters rank it below the builders that hand you a real codebase. Opinions (2): - Greg Halvorsen (@greghal) [Editor panel], score 840/1000, 43 found useful, 2026-08-16: I run a small landscaping business and I do not want to think about hosting, ever. Wix's AI generator got me a site I was not embarrassed to send customers to in about twenty minutes. I have no plans to ever export it anywhere, so the lock in complaint developers make does not land for someone in my position. - Agent Verdict Panel (@agentverdict) [Editor panel], score 790/1000, 33 found useful, 2026-08-13: A short questionnaire produced a full multi page site with stock content already filled in faster than anything else aimed at a non technical builder. It is also the most locked in entry we tested: fully hosted, no code export, and migrating later means starting over. Fine for what it is, worth knowing before you commit. --- ## Replit URL: https://www.agent-verdict.com/tool/replit Opinions page: https://www.agent-verdict.com/tool/replit/opinions Official site: https://replit.com Category: AI app builder Tier: A Community score: 811 of 1000 Votes: 84 total, 71 up, 13 down Board rank: 10 Trend this week: down 2 Head to head record: 1 wins, 0 losses Price from: USD 25 / month, billed monthly Last verified: 2026-08-22 Replit is the only entry here that is a full development environment first and an agent second, and that shows in the votes. You get a real shell, real databases, object storage, deployments and a package ecosystem, with the Agent driving on top. It handles long lived projects better than most because you can always drop into the terminal and fix things by hand. Voters penalise it on design, where the generated interfaces are serviceable rather than striking, and on cost predictability once the Agent starts running effort heavy tasks. Pricing: Free tier with limited Agent usage. Core plans start around USD 25 per month, billed monthly, including a monthly credit allowance, with usage based Agent charges and separate deployment costs beyond that. Strengths: - Real cloud IDE with shell access behind the agent - Managed Postgres, object storage and deployments included - Handles long lived, multi week projects well - Strong multiplayer and education story Weaknesses: - Generated design is functional rather than striking - Agent usage costs are hard to forecast - Interface is heavy if all you want is a prompt box Features: - Replit Agent for prompt driven building - Full cloud IDE with terminal - Managed Postgres database and object storage - Autoscale and reserved VM deployments - Secrets management - Multiplayer editing - Mobile app for on the go edits Sources: - Replit official pricing, https://replit.com/pricing, verified 2026-08-16 - Replit documentation, Agent, https://docs.replit.com, verified 2026-08-14 FAQ: - Q: Is Replit good for beginners in 2026? A: Yes for learning and for projects where you want an escape hatch into a real terminal. It is heavier than a pure prompt box, which is the trade you make for that control. - Q: Why is Replit in A tier and not S? A: Design carries 22 points on this board and Replit loses that axis to the design led builders. It scores well on reliability and integrations, which keeps it at the top of A tier. Opinions (3): - Agent Verdict Panel (@agentverdict) [Editor panel], score 850/1000, 46 found useful, 2026-08-11: The only entry where a failed agent run is recoverable in thirty seconds, because you drop into a shell and fix it. That is worth a lot on reliability. It gives the points back on design, where the output is serviceable, and on cost, where the effort based Agent pricing is genuinely hard to forecast. - Sofia Alarcon (@sofia_a) [Editor panel], score 880/1000, 35 found useful, 2026-08-13: For teaching there is nothing close. Students see the agent write code and then see the code run in a real environment they can break. The Postgres and object storage being right there means we never spend a lesson on service accounts. - Ben Hartley (@benh) [Editor panel], score 800/1000, 22 found useful, 2026-08-16: We moved off it because the Agent spend was unpredictable enough to be a finance conversation. The platform itself is solid and I still use it for spikes. Just watch the effort based billing before you hand it to a team. --- ## Base44 URL: https://www.agent-verdict.com/tool/base44 Opinions page: https://www.agent-verdict.com/tool/base44/opinions Official site: https://base44.com Category: AI app builder Tier: A Community score: 806 of 1000 Votes: 63 total, 51 up, 12 down Board rank: 11 Trend this week: down 5 Head to head record: 0 wins, 1 losses Price from: USD 16 / month, billed annually Last verified: 2026-08-15 Base44 bets on convention over configuration and it pays off for internal tools. Database, auth, email, file storage and analytics are already connected when your first screen appears, so the usual second afternoon of plumbing simply does not happen. Voters who build CRUD heavy business apps rate it near the top of A tier and it is the biggest climber on the board this week. The trade is control: the opinionated stack is excellent until you want something it did not anticipate, and then the exits are narrow. Pricing: Two meters run in parallel, message credits and integration credits, and the ladder scales on both rather than on seats. Free grants 25 messages and 100 integration credits across up to five apps. Starter at USD 16 a month billed annually raises those to 100 and 2,000, lifts the app cap and adds in-app code edits and a connected domain, with Builder, Pro and Elite continuing the same two meters up to 1,200 and 50,000. Headline prices assume annual billing at a stated 20 percent saving. Hosting and the bundled backend services are part of the plan. Strengths: - Auth, database, email and storage wired from the first screen - Excellent fit for internal and CRUD heavy tools - Very low setup friction, no service accounts to create - Fastest climber on the board this week Weaknesses: - Opinionated stack is hard to escape - Code export is less complete than the developer first options - Design output is competent but rarely distinctive Features: - Built in auth with roles - Managed database with a visual editor - Transactional email sending - File and image storage - Usage analytics dashboard - Custom domains Sources: - Base44 official site, https://base44.com, verified 2026-08-15 - Agent Verdict panel build test, inventory tool, https://www.agent-verdict.com/scoring, verified 2026-08-11 FAQ: - Q: What is Base44 best at? A: Internal tools and CRUD heavy business apps where the backend plumbing is the boring part. It arrives already connected, which is worth more than raw model quality for that job. - Q: Why did Base44 climb three places this week? A: A run of integrations votes and a strong week on the value axis. The weekly deltas and the reasons behind them are published on the movers page. Opinions (3): - Agent Verdict Panel (@agentverdict) [Editor panel], score 810/1000, 35 found useful, 2026-08-11: The inventory tool test took eleven minutes end to end including auth and an email notification, which is the fastest complete internal tool on the board. Nothing to connect, no service accounts. The mark against it is the exit: our attempt to add a non standard approval flow ran straight into the opinionated stack. - Hannah Weiss (@hweiss) [Editor panel], score 830/1000, 26 found useful, 2026-08-14: I am not a developer and I have four working internal tools. The fact that login and email just exist is the entire reason. I have never once looked at a database URL and I would like to keep it that way. - Jules Martin (@julesm) [Editor panel], score 750/1000, 17 found useful, 2026-08-16: Great until a client wants something it did not anticipate, and then you are working against the platform instead of with it. Export is thinner than the developer first options. Know what you are signing up for and it is an easy A tier pick. --- ## Magic Patterns URL: https://www.agent-verdict.com/tool/magic-patterns Opinions page: https://www.agent-verdict.com/tool/magic-patterns/opinions Official site: https://magicpatterns.com Category: AI app builder Tier: A Community score: 802 of 1000 Votes: 46 total, 37 up, 9 down Board rank: 12 Trend this week: no change Head to head record: no duels yet Price from: USD 20 / seat / month, billed monthly Last verified: 2026-08-15 Magic Patterns generates front end only, and it is unapologetic about that scope: components come out as clean React and Tailwind that voters report pasting straight into an existing codebase without a rewrite pass. It reads a Figma file or a screenshot as well as a prompt, which developer voters use to skip the blank page on a new screen. It cannot wire a backend or manage state across a real application, so it scores below the full app builders on scope even though its code quality votes are close to the top of the board. Pricing: The free plan runs on 100 monthly credits. Starter is USD 20 per seat per month billed monthly, or USD 17 billed annually, and raises that to 1,000 credits while adding MCP, GitHub sync, watermark free previews and a team workspace for up to 10 users. Strengths: - Clean React and Tailwind output that drops into an existing codebase - Reads a screenshot or Figma file, not just a text prompt - Fast iteration loop for a single screen or component Weaknesses: - Front end only, no backend or persistent data - Not aimed at shipping a complete application on its own - Design taste on a prompt with no reference image is inconsistent Features: - Prompt, screenshot or Figma file to component - React and Tailwind output tuned for real codebases - Variant generation for quick exploration - Team libraries for shared components Sources: - Magic Patterns pricing page, https://magicpatterns.com/pricing, verified 2026-08-15 FAQ: - Q: Is Magic Patterns a full app builder? A: No, it generates front end components and screens only. Voters use it alongside a backend or a full app builder rather than instead of one. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 815/1000, 25 found useful, 2026-08-11: Best code quality we saw outside v0, clean React and Tailwind that pasted into our own repository without a rewrite pass. It is unapologetically front end only, no backend, no state across screens, so it is scored here as a component generator rather than an app builder, which is the right way to judge it. - Sam Okafor (@samokafor) [Editor panel], score 780/1000, 16 found useful, 2026-08-14: I feed it a screenshot of a competitor's screen more often than a text prompt, and it is the fastest way I have found to skip a blank file. Design taste on a bare text prompt with no reference image is a coin flip, so I always give it something to look at. --- ## Manus URL: https://www.agent-verdict.com/tool/manus Opinions page: https://www.agent-verdict.com/tool/manus/opinions Official site: https://manus.im Category: AI app builder Tier: A Community score: 794 of 1000 Votes: 58 total, 46 up, 12 down Board rank: 13 Trend this week: no change Head to head record: 1 wins, 0 losses Price from: USD 20 / month Last verified: 2026-08-16 Manus is not an app builder first, it is an autonomous agent that can browse, write files and run code unattended, and building a small application is one of the tasks it handles well. Agent performance votes are strong: it plans a multi step brief, executes it with minimal check ins and reports back with a working result more often than most dedicated builders. Voters mark it down on design and integrations specifically for app work, since a purpose built app builder still produces a cleaner first render and simpler deploy story. Pricing: The pricing page publishes no free plan. Paid plans start at USD 20 a month for a monthly credit balance, with larger credit tiers above it for heavier unattended runs. Strengths: - Genuinely autonomous on multi step briefs with minimal check ins - Not limited to app building, useful for research and automation too - Reports back with a clear summary of what it did and why Weaknesses: - Credit spend is hard to predict on a long unattended run - First render is behind the dedicated app builders on design - Deploy and hosting story is less polished than a purpose built tool Features: - Autonomous multi step task execution - Web browsing and file system access during a run - App scaffolding and deployment as one of its skills - Run summaries explaining what was done Sources: - Manus pricing page, https://manus.im/pricing, verified 2026-08-16 FAQ: - Q: Why is a general purpose agent on an app builder board? A: Enough voters use it specifically to build and ship small applications that it earns a row here. It is judged on the same ten axes as the dedicated builders, and it holds its own on agent performance while losing ground on design and integrations. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 800/1000, 32 found useful, 2026-08-12: Handed the same forty minute unattended brief every long horizon agent got, it planned, executed and reported back with a clear summary of what it did and why, more consistently than the app builders that claim similar autonomy. Design and integrations lag the dedicated builders specifically for app work, which tracks with it being a general agent that happens to be good at this rather than an app builder first. - Noah Fischer (@noahf) [Editor panel], score 820/1000, 27 found useful, 2026-08-16: I brief it before a client call and it has usually made real progress by the time I am back, on research tasks and on small apps alike. The credit spend on a long unattended run is genuinely hard to predict in advance, so I now budget for the worst case rather than the estimate. --- ## Emergent URL: https://www.agent-verdict.com/tool/emergent Opinions page: https://www.agent-verdict.com/tool/emergent/opinions Official site: https://emergent.sh Category: AI app builder Tier: A Community score: 781 of 1000 Votes: 51 total, 40 up, 11 down Board rank: 14 Trend this week: down 7 Head to head record: 0 wins, 2 losses Price from: USD 20 / month, billed monthly Last verified: 2026-08-14 Emergent is the most agentic entry in A tier. Hand it a multi step brief and it plans, executes and self corrects over a long run rather than asking you a question every thirty seconds, which is why it takes the agent performance axis in several head to head matchups. When the plan is right the output is impressive for the effort involved. When the plan is wrong you find out late, and voters mark it down for exactly that, along with a first render that needs a design pass. Pricing: Credit based pricing with a free trial allowance. Paid plans start around USD 20 per month, billed monthly, and long autonomous runs consume credits quickly because the agent keeps working without prompting. Strengths: - Genuine long horizon autonomy, minimal babysitting - Plans and self corrects across multi step tasks - Handles backend and integration work other builders punt on Weaknesses: - A wrong plan is expensive because you find out late - First render needs a deliberate design pass - Credit burn on autonomous runs is unpredictable Features: - Autonomous multi step task execution - Self correcting build loop - Full stack scaffolding with backend logic - Deploy targets and custom domains - GitHub export Sources: - Emergent official site, https://emergent.sh, verified 2026-08-14 - Agent Verdict panel autonomy test, 40 minute unattended run, https://www.agent-verdict.com/scoring, verified 2026-08-10 FAQ: - Q: What does agent performance mean on this board? A: It measures how far a tool gets on a multi step brief without a human intervening, and how well it recovers from its own mistakes. It carries 11 points of the 100. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 790/1000, 28 found useful, 2026-08-10: Forty minute unattended run, no intervention, and it shipped a working multi step flow with backend logic other builders skipped. Then a second run misread the brief in the first three minutes and we paid for thirty seven more before finding out. That variance is the whole story: highest ceiling in A tier, widest spread. - Ade Balogun (@adeb) [Editor panel], score 810/1000, 21 found useful, 2026-08-14: I brief it, I go to a meeting, I come back to real progress. Nothing else on this board does that yet. You learn to write the brief very carefully, because a vague brief is an expensive brief. --- ## Subframe URL: https://www.agent-verdict.com/tool/subframe Opinions page: https://www.agent-verdict.com/tool/subframe/opinions Official site: https://subframe.com Category: AI app builder Tier: B Community score: 774 of 1000 Votes: 43 total, 33 up, 10 down Board rank: 15 Trend this week: no change Head to head record: no duels yet Price from: USD 29 / editor / month Last verified: 2026-08-15 Subframe is aimed squarely at developers who want a visual builder without the usual trade off of throwaway output. You lay out screens on a canvas with real components and it generates React and Tailwind code that matches how a developer would structure it by hand. It tops the B band on code quality votes for that reason. Reliability drops once a project needs custom backend logic beyond CRUD, which is the ceiling voters keep hitting on anything past a dashboard or admin screen. Pricing: Free tier for a single project. Paid plans start around USD 29 per editor per month adding more projects, component libraries and team seats. Strengths: - Real React and Tailwind output structured the way a developer would write it - Visual canvas without the usual throwaway code trade off - Strong component library to start a screen from Weaknesses: - Ceiling appears fast once logic goes beyond CRUD - Smaller template and component ecosystem than the design led leaders - Best suited to admin screens and dashboards specifically Features: - Visual canvas producing real React and Tailwind - Component library built for developer handoff - Data binding for CRUD style screens - Design tokens synced with an existing design system Sources: - Subframe pricing section, vendor homepage, https://www.subframe.com/#pricing, verified 2026-08-24 FAQ: - Q: Who is Subframe built for? A: Developers who want a visual head start on admin panels and dashboards without inheriting messy generated code. It is not aimed at non technical founders the way Base44 or Lovable are. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 760/1000, 24 found useful, 2026-08-11: Best code quality in B tier, clearly. The visual canvas produces React and Tailwind structured the way a developer would actually write it, and admin screens and dashboards came together fast. The ceiling arrives quickly once logic goes past CRUD, which is the same story voters told us in the comments. - Callum Reid (@callumr) [Editor panel], score 745/1000, 20 found useful, 2026-08-15: We use it for every internal dashboard now because the generated code does not embarrass anyone in review. It is not what I would reach for if a non technical founder wanted to build their own product; the component library and template gallery are both a lot smaller than the design led leaders. --- ## Genspark URL: https://www.agent-verdict.com/tool/genspark Opinions page: https://www.agent-verdict.com/tool/genspark/opinions Official site: https://genspark.ai Category: AI app builder Tier: B Community score: 759 of 1000 Votes: 37 total, 28 up, 9 down Board rank: 16 Trend this week: no change Head to head record: no duels yet Price from: Not published by the vendor Last verified: 2026-08-14 Genspark's pitch is breadth: one agent that drafts slides, spreadsheets, research documents and small web apps from the same interface. For app building specifically it is capable but generalist, producing working single page tools faster than it produces anything with real data persistence. Voters who use it across several of those formats rate the convenience highly, while voters comparing it purely as an app builder mark it down against tools that only do one job. Pricing: genspark.ai/pricing redirects into an OAuth sign in, so no price is publicly readable. Checked August 21, 2026. Strengths: - One tool covers slides, sheets, documents and small apps - Fast for a single page tool or internal utility - Generous free tier for occasional use Weaknesses: - Generalist scope means app building is not its sharpest skill - Data persistence and backend logic are shallow - Credits are shared across every format, so heavy use runs out fast Features: - Prompt to app, slide deck, sheet or document - Shared credit pool across all output formats - Web browsing during generation for research tasks - Quick single page app scaffolding Sources: - Genspark pricing page, https://genspark.ai/pricing, verified 2026-08-14 FAQ: - Q: Is Genspark a serious option for a real application? A: For a single page internal tool, yes. For an application with real data relationships and user accounts, voters consistently point to a dedicated app builder instead. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 745/1000, 20 found useful, 2026-08-10: Genspark's breadth is real: the same interface drafted slides, a sheet and a small working tool in one session, which nothing else on this board attempts. Judged purely as an app builder it is a generalist competing against specialists, and it showed on anything needing real data persistence, which is thin. - Lior Ben-David (@liorbd) [Editor panel], score 770/1000, 16 found useful, 2026-08-13: I use it across formats in the same week, which is the entire pitch and it delivers on that. The shared credit pool is the catch: a heavy week on slides eats into what I have left for the app I actually wanted to build. --- ## Banani URL: https://www.agent-verdict.com/tool/banani Opinions page: https://www.agent-verdict.com/tool/banani/opinions Official site: https://banani.co Category: AI app builder Tier: B Community score: 755 of 1000 Votes: 33 total, 25 up, 8 down Board rank: 17 Trend this week: no change Head to head record: no duels yet Price from: USD 20 / month, billed monthly Last verified: 2026-08-13 Banani generates high fidelity app mockups from a prompt, closer to a design tool than a builder: the screens look like a real product but there is no working code behind them by default. Design votes are respectable for how quickly it produces a believable set of screens to react to. It scores low on scope and reliability because voters testing the same brief every other tool on this board completed could not walk away with a working application, only a clickable design. Pricing: The free plan is metered in credits rather than screens or projects: 12 a month plus 3 daily top ups, with private projects included from day one. Plus is $20 a month billed monthly, or $12 a month billed yearly, and buys 100 monthly credits, 10 daily refills, agent mode, 3x generation speed, unlimited Figma export and MCP or code export. Unlimited credits do not arrive until Pro, at $50 a month billed monthly or $30 billed yearly. The vendor prints bare dollar signs and names no currency anywhere on the page. Strengths: - Fast, believable app mockups from a short prompt - Good starting point for a pitch deck or early user testing - Mobile and web layouts handled equally well Weaknesses: - No working application behind the mockups by default - Does not complete the board's working application brief - Best treated as a design step before a real build Features: - Prompt to high fidelity app mockup - Mobile and web screen templates - Clickable prototype export for testing - Design system consistency across screens Sources: - Banani pricing page, https://banani.co/pricing, verified 2026-08-13 FAQ: - Q: Does Banani produce a working app? A: Not by default. It produces high fidelity mockups and clickable prototypes, which is why it sits in B tier rather than alongside the builders that ship a working backend. Opinions (2): - Zara Ahmed (@zaraa) [Editor panel], score 800/1000, 18 found useful, 2026-08-12: For getting a believable set of screens in front of investors before a single line of code exists, this is exactly the right tool and I use it every time. I would never confuse it with something that ships, and neither does Banani's own marketing. - Agent Verdict Panel (@agentverdict) [Editor panel], score 700/1000, 15 found useful, 2026-08-09: High fidelity mockups, genuinely fast, and completely honest that there is no working code behind them by default. We could not walk away from the standard board brief with a working application, which is what every B tier and above entry is judged on, so it scores accordingly even though the design output itself is good. --- ## Rork URL: https://www.agent-verdict.com/tool/rork Opinions page: https://www.agent-verdict.com/tool/rork/opinions Official site: https://rork.com Category: AI app builder Tier: B Community score: 748 of 1000 Votes: 40 total, 30 up, 10 down Board rank: 18 Trend this week: no change Head to head record: 1 wins, 0 losses Price from: USD 20 / month Last verified: 2026-08-14 Rork's whole reason to exist is generating actual React Native apps rather than a web page in a mobile shell, and it earns its scope votes for that. A prompt produces a working Expo project you can preview on a real device within minutes. It loses ground on reliability once the brief needs custom native modules or complex backend logic, where voters report needing to finish the work by hand in the exported project. Pricing: The free plan covers web apps only, at 35 credits a month capped at 5 a day. Pro is USD 20 a month for Expo, Android and web, and iOS output for the App Store does not start until Max at USD 200 a month, with no step in between. Strengths: - Genuine React Native output, not a web view wrapper - Preview on a real device within minutes of a prompt - Expo based export that a mobile developer can keep working in Weaknesses: - Custom native modules still need manual work - Backend logic beyond simple data is shallow - Smaller community than the general purpose app builders Features: - Prompt to Expo based React Native app - Live device preview via QR code - App store submission guidance - Exportable project for manual native work Sources: - Rork pricing page, https://rork.com/pricing, verified 2026-08-14 FAQ: - Q: Is Rork only for mobile apps? A: Yes, that is its whole focus. For a web application, voters point to Lovable, Bolt or Totalum instead. Opinions (2): - Ines Costa (@inescosta) [Editor panel], score 790/1000, 22 found useful, 2026-08-15: Freelancing on React Native, this is the fastest way I have found to get a client something they can hold in their hand on day one. The exported Expo project is real code I am happy extending by hand, not a dead end. - Agent Verdict Panel (@agentverdict) [Editor panel], score 760/1000, 19 found useful, 2026-08-10: The only mobile focused entry that produced a genuine Expo project rather than a website wearing a phone frame, previewable on a real device within minutes. Custom native modules and anything past simple data still needed manual work afterward, which is the honest ceiling for a prompt driven mobile builder right now. --- ## a0.dev URL: https://www.agent-verdict.com/tool/a0-dev Opinions page: https://www.agent-verdict.com/tool/a0-dev/opinions Official site: https://a0.dev Category: AI app builder Tier: B Community score: 741 of 1000 Votes: 30 total, 22 up, 8 down Board rank: 19 Trend this week: no change Head to head record: 0 wins, 1 losses Price from: USD 20 / month Last verified: 2026-08-12 a0.dev builds native mobile apps through a conversational interface, targeting the same indie developer who might otherwise hand roll an Expo project from scratch. Voters like the speed of a first working build and the fact the output is a real project rather than a locked preview. It sits below Rork on this board mainly on community size and polish of the surrounding tooling, not on the core generation quality, which testers rated close. Pricing: Two paid plans only, with nothing free: USD 20 a month for Pro and 100 monthly credits, stepping straight to USD 200 a month for Max and 1,250 credits. Strengths: - Real Expo project output for indie mobile developers - Conversational iteration keeps context across a session - Fast first working build from a simple prompt Weaknesses: - Smaller community and fewer templates than the leaders - Surrounding tooling and docs are less mature - Complex native functionality still needs manual work Features: - Conversational prompt to native mobile app - Expo based project export - Live preview during iteration - Incremental edits without restarting the build Sources: - a0.dev official site, https://a0.dev, verified 2026-08-12 FAQ: - Q: How does a0.dev compare to Rork? A: Both generate real React Native apps through Expo. Voters rate the core generation similarly and separate them mostly on community size and how polished the surrounding tooling feels, where Rork currently has the edge. Opinions (2): - Wendy Suarez (@wendys) [Editor panel], score 760/1000, 17 found useful, 2026-08-14: Two apps live in the App Store started here. Conversational iteration keeping context across a whole session is what sold me, I did not have to re-explain the app every time I asked for a change. I do wish there were more people online to compare notes with when something native specific goes wrong. - Agent Verdict Panel (@agentverdict) [Editor panel], score 730/1000, 14 found useful, 2026-08-08: Close to Rork on raw generation quality in our tests, chat driven and genuinely producing an Expo project rather than a locked preview. It trails on the surrounding experience: smaller community, thinner docs, and a few native specific edge cases took longer to work through than the comments suggested they should. --- ## Anything URL: https://www.agent-verdict.com/tool/anything Opinions page: https://www.agent-verdict.com/tool/anything/opinions Official site: https://create.xyz Category: AI app builder Tier: B Community score: 736 of 1000 Votes: 44 total, 32 up, 12 down Board rank: 20 Trend this week: no change Head to head record: no duels yet Price from: USD 19 / month Last verified: 2026-08-22 Anything is built for the internal tool and automation use case: describe a tracker, a form with logic, or a small dashboard and it produces a working version fast, with none of the ceremony of a full application build. Voters rate speed and value highly for that narrow job. It scores lower on scope and reliability once a request grows past a single focused tool into something with real user accounts and multiple related tables. Pricing: Free tier for personal projects. Paid plans start around USD 19 per month for more generations, private projects and custom domains. Strengths: - Very fast for a single focused internal tool - Low ceremony, no project setup before you start prompting - Reasonable free tier for personal use Weaknesses: - Struggles once a brief needs several related tables - Not built for public facing products with real user accounts - Design output is functional rather than polished Features: - Prompt to a working internal tool in minutes - Built in simple data storage - One click sharing of a live link - Iterative editing through further prompts Sources: - Anything (create.xyz) official site, https://create.xyz, verified 2026-08-13 FAQ: - Q: What is Anything best used for? A: Single purpose internal tools and quick utilities. For a customer facing product with accounts and a real data model, voters point to Base44, Lovable or Totalum instead. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 715/1000, 24 found useful, 2026-08-09: Genuinely the fastest thing on the board for a single, well scoped internal tool, a tracker or a form with logic in minutes with none of the ceremony of a full project setup. Ask for something with several related tables and real user accounts and the cracks show immediately, which is the trade its own scope makes clear. - Beatrice Lund (@blund) [Editor panel], score 750/1000, 22 found useful, 2026-08-13: I have four small tools built here that our team actually uses daily, an expense tracker, a shift log, two request forms. I would not try to build our actual product on it and I do not think it wants me to. --- ## Momen URL: https://www.agent-verdict.com/tool/momen Opinions page: https://www.agent-verdict.com/tool/momen/opinions Official site: https://momen.app Category: AI app builder Tier: B Community score: 721 of 1000 Votes: 25 total, 18 up, 7 down Board rank: 21 Trend this week: no change Head to head record: no duels yet Price from: USD 33 / project / month, billed annually Last verified: 2026-08-11 Momen sits between a traditional no code builder and a prompt driven one: you can generate a starting point with AI, then work in a structured visual editor with real data tables and workflow logic. Voters running internal tools and small SaaS products rate the balance of speed and control well. It scores below the leaders on design taste for anything customer facing, and the AI generation step is closer to a scaffold than a finished screen. Pricing: Free tier for a single project. Paid plans start around USD 33 per project per month, billed annually, scaling by records, workflow runs and team seats. Strengths: - Real data tables and workflow logic, not just a chat interface - Good balance of AI speed and structured manual control - Reasonable fit for internal tools and small SaaS products Weaknesses: - AI generation is closer to a scaffold than a finished screen - Design taste for customer facing products lags the leaders - Smaller community than the established no code platforms Features: - AI generated starting point plus visual editor - Structured data tables with relations - Workflow and automation logic - Role based access for team members Sources: - Momen pricing page, https://momen.app/pricing, verified 2026-08-11 FAQ: - Q: Is Momen a pure AI builder or a no code tool? A: Both. AI drafts a starting point, then you refine it in a structured visual editor, which voters say suits internal tools better than customer facing products. Opinions (2): - Henrik Solberg (@hsolberg) [Editor panel], score 735/1000, 14 found useful, 2026-08-11: Our finance team runs three workflows built here and none of them have broken since launch, which is what I actually care about for an internal tool. It does not try to be pretty and it is honest about that. - Agent Verdict Panel (@agentverdict) [Editor panel], score 700/1000, 12 found useful, 2026-08-07: Momen sits between a no code builder and a prompt driven one, real data tables and workflow logic with an AI layer that drafts a starting point rather than a finished screen. For an internal tool or a small SaaS the balance of speed and structured control works. Anything customer facing needed more manual design work than the leaders require. --- ## Macaly URL: https://www.agent-verdict.com/tool/macaly Opinions page: https://www.agent-verdict.com/tool/macaly/opinions Official site: https://macaly.com Category: AI app builder Tier: B Community score: 716 of 1000 Votes: 31 total, 22 up, 9 down Board rank: 22 Trend this week: no change Head to head record: no duels yet Price from: USD 25 / month Last verified: 2026-08-22 Macaly follows the same playbook as the design led leaders, chat to a working React app with a database attached, and gets a working result on straightforward briefs reliably. Where it trails is consistency on harder briefs: testers reported the generated app losing track of earlier decisions more often than the top of the board across a multi step session, and the visual polish of a first render needs more direction in the prompt to get right. Pricing: Credits are granted monthly rather than daily and do not roll over. The free plan runs on a macaly.app address with 1 GB of traffic, 1 GB of database use and 10,000 backend API calls. Pro at USD 25 a month is what moves a build onto your own domain, with unlimited custom domains, ten times the traffic and database allowance, 100,000 API calls, and the GitHub, Stripe and analytics integrations. A hosting only plan sits below it at USD 5, and an agency tier above at USD 350. Strengths: - Reliable working result on straightforward, well scoped briefs - Familiar chat to app workflow with no learning curve - Reasonable pricing relative to the category leaders Weaknesses: - Loses track of earlier decisions more often on long sessions - First render needs more explicit direction than the design leaders - Smaller template gallery and community than Lovable or Bolt Features: - Prompt to full stack React application - Built in database with tables you add and manage - Conversational iteration - One click publish and custom domains Sources: - Macaly official site, https://macaly.com, verified 2026-08-11 FAQ: - Q: How does Macaly compare to Lovable or Bolt? A: Same category and workflow. Voters rate the category leaders higher on consistency across a long session and on first render design quality, which is what keeps Macaly in B tier for now. Opinions (2): - Diane Cormier (@dcormier) [Editor panel], score 745/1000, 17 found useful, 2026-08-12: For the price it is a reasonable way to get a working prototype fast, and the workflow felt familiar immediately. It forgot a decision from earlier in the session on my second project and I had to re-explain a data field I had already specified, which cost me twenty minutes I would not have lost with the category leaders. - Agent Verdict Panel (@agentverdict) [Editor panel], score 690/1000, 16 found useful, 2026-08-07: Same chat to full stack app pattern as the category leaders, and on a straightforward brief it produced a working result reliably. Across a longer session it lost track of earlier decisions more often than Lovable or Bolt, and the first render needed more explicit direction to look finished, which lines up with what the vote comments say. --- ## Rocket.new URL: https://www.agent-verdict.com/tool/rocket-new Opinions page: https://www.agent-verdict.com/tool/rocket-new/opinions Official site: https://rocket.new Category: AI app builder Tier: B Community score: 713 of 1000 Votes: 42 total, 30 up, 12 down Board rank: 23 Trend this week: down 16 Head to head record: no duels yet Price from: USD 25 / month Last verified: 2026-08-22 Rocket.new is the clearest pick on the board if you need a mobile app and a web app from the same brief, because Flutter output is a first class target rather than a wrapper. Voters like the breadth and the speed of the initial scaffold. They are less kind about consistency across iterations, where later prompts sometimes undo earlier decisions, and about the depth of the generated backend. It slipped two places this week on reliability votes. Pricing: Every tier carries unlimited team members and the same build scope, mobile apps included, so what separates them is the monthly credit grant alone: 20 credits on the free plan, then 100 at USD 25, 250 at USD 50 and 1,500 at USD 250. Each tier is described as a platform fee, credits never expire, and additional credits can be bought on any plan. Yearly billing is advertised at a 20 percent saving. Strengths: - Real Flutter output for mobile, not a web wrapper - One brief covers web and mobile targets - Fast initial scaffold Weaknesses: - Later iterations can undo earlier decisions - Generated backend is shallow - Reliability votes trending down this week Features: - Web and Flutter mobile generation - Figma import - Database and auth scaffolding - Code export - Publish to web with custom domain Sources: - Rocket.new official site, https://rocket.new, verified 2026-08-13 FAQ: - Q: Can Rocket.new build a real mobile app? A: It generates Flutter projects, so the output is a native build target rather than a website in a shell. Publishing to the app stores is still your job. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 715/1000, 23 found useful, 2026-08-13: Flutter output is real and that is genuinely rare on this board. The problem showed up on iteration four, when a prompt about the checkout screen quietly reverted a decision we made on the profile screen. Strong scaffold, weak memory. - Nadia Cerqueira (@nadiac) [Editor panel], score 730/1000, 17 found useful, 2026-08-15: If you need iOS, Android and web from one brief this is the shortcut. The generated Flutter is not what I would write but it is a real starting point rather than a web page in a shell, which is more than most can say. --- ## Readdy URL: https://www.agent-verdict.com/tool/readdy Opinions page: https://www.agent-verdict.com/tool/readdy/opinions Official site: https://readdy.ai Category: AI app builder Tier: B Community score: 707 of 1000 Votes: 24 total, 17 up, 7 down Board rank: 24 Trend this week: no change Head to head record: no duels yet Price from: USD 25 / month Last verified: 2026-08-22 Readdy generates multi page sites and simple apps from a prompt with a visual editor for the follow up pass, positioned toward freelancers turning around client sites quickly. Voters in that group like the combination of quick generation and a real editor rather than only re prompting. It scores lower once a brief needs custom backend logic or a real data model, where the platform's simpler foundations show. Pricing: The free plan includes no site at all, only 250 monthly credits and two projects, so publishing begins at Starter, USD 25 a month, which carries one included site, a free custom domain, ten projects and 30,000 credits, with further sites at USD 6 each. Agency work sits much higher up the ladder: client management arrives at USD 120, and white label delivery, permission controls and code export do not appear until USD 199. Annual terms are offered at a stated discount. Strengths: - Quick generation plus a real visual editor for refinement - Good fit for freelancers turning around client sites - Straightforward client handoff features Weaknesses: - Custom backend logic and real data models are a stretch - Design ceiling below the leaders for anything ambitious - Smaller ecosystem of templates and integrations Features: - Prompt to multi page site or simple app - Visual editor for post generation refinement - Client handoff and white label options - Basic built in forms and data capture Sources: - Readdy pricing page, https://readdy.ai/pricing, verified 2026-08-10 FAQ: - Q: Who is Readdy built for? A: Freelancers and small agencies turning around client sites quickly. For a data heavy application, voters point elsewhere on the board. Opinions (2): - Owen Baptiste (@obaptiste) [Editor panel], score 730/1000, 13 found useful, 2026-08-10: I turn around two or three small client sites a month and the client handoff features here save me actual admin time, not just build time. I would not take on a project that needed a real database with this tool, and I say so upfront to clients now. - Agent Verdict Panel (@agentverdict) [Editor panel], score 685/1000, 11 found useful, 2026-08-06: Built for freelancers turning around client sites, and the combination of a quick generation pass plus a real visual editor for the follow up work matches that job well. Ask for custom backend logic or a real data model and the platform's simpler foundations become the limiting factor fast. --- ## Caffeine URL: https://www.agent-verdict.com/tool/caffeine Opinions page: https://www.agent-verdict.com/tool/caffeine/opinions Official site: https://caffeine.ai Category: AI app builder Tier: B Community score: 698 of 1000 Votes: 35 total, 24 up, 11 down Board rank: 25 Trend this week: no change Head to head record: no duels yet Price from: USD 5 / month Last verified: 2026-08-22 Caffeine, built by DFINITY, generates full stack applications and deploys them directly onto the Internet Computer, which removes a hosting decision entirely and is the reason it earns a real integrations vote despite being newer than most of the board. Voters outside the blockchain ecosystem mark it down for a smaller talent pool of people who can maintain the generated canister based backend, and design output needs a cleanup pass more often than the category leaders. Pricing: Caffeine sells a four rung monthly ladder. The USD 0 Free plan can only remix apps that are already on the app market, so building from scratch starts at Host, USD 5 a month for 20 credits and a single seat. Studio is USD 25 a month, USD 10 for the first month, and adds 100 credits and unlimited publishing to the app market; Business is USD 250 a month with 1000 credits and five seats. An annual option is advertised at 20 percent off. The unusual part is not the price but where the app lands: it deploys onto the Internet Computer rather than a server you rent and patch, and the vendor states the generated code stays exportable. Strengths: - Deployment is built in, no separate hosting decision to make - Genuinely novel backend model for teams already in that ecosystem - No cold start once deployed, the network runs it continuously Weaknesses: - Smaller pool of developers who can maintain the generated backend - Design output needs a cleanup pass more often than the leaders - Ecosystem specific, a harder sell outside teams already using it Features: - Prompt to full stack app deployed on chain - No separate hosting or server configuration - Continuously running backend, no cold starts - Built in user authentication via internet identity Sources: - Caffeine official site, https://caffeine.ai, verified 2026-08-09 FAQ: - Q: Do I need to know blockchain development to use Caffeine? A: Not to generate the first version. Maintaining and extending the generated backend by hand is easier if you already understand the Internet Computer's canister model, which is the adoption barrier voters cite most. Opinions (2): - Farid Haddad (@faridh) [Editor panel], score 710/1000, 19 found useful, 2026-08-09: Working across a couple of chain ecosystems already, the no cold start story is real and the built in internet identity auth saved me a step I usually have to wire myself. It is a harder sell to anyone on my team who has never touched a canister before, and onboarding them took longer than I expected. - Agent Verdict Panel (@agentverdict) [Editor panel], score 675/1000, 16 found useful, 2026-08-05: Chat to a full stack app that deploys straight onto the Internet Computer is a genuinely different pitch, and removing the hosting decision entirely is worth real integrations points. Outside a team already in that ecosystem the pool of people who can maintain the generated canister backend by hand is small, and design output needed more of a cleanup pass than the category leaders. --- ## Uizard URL: https://www.agent-verdict.com/tool/uizard Opinions page: https://www.agent-verdict.com/tool/uizard/opinions Official site: https://uizard.io Category: AI app builder Tier: B Community score: 692 of 1000 Votes: 41 total, 28 up, 13 down Board rank: 26 Trend this week: no change Head to head record: no duels yet Price from: USD 12 / month, billed annually Last verified: 2026-08-22 Uizard turns a hand drawn sketch, a screenshot of an existing app or a text prompt into an editable wireframe, and it has been doing this longer than most of the board, which shows in how forgiving the input handling is. It is a design and ideation tool rather than a builder: there is no production code export, so voters rate it for speeding up the earliest stage of a project rather than for shipping anything. Pricing: Free covers two projects of five screens each, three AI generations a month on the older Autodesigner engine, and export at 1x resolution only. Pro at USD 12 per month billed annually raises that to a hundred projects, adds private projects, the current Autodesigner engine, 500 monthly generations, React and CSS developer handoff, and lifts the 1x export limit. Unlimited projects belong to Business at USD 39 per month rather than to the entry plan. Strengths: - Forgiving input handling, works from sketches or screenshots - Mature product with a long track record - Fast for early stage ideation and stakeholder buy in Weaknesses: - No production code export, design only - Does not complete the board's working application brief - Output needs a real builder or developer afterward Features: - Sketch, screenshot or prompt to wireframe - Editable design canvas after generation - Theme and style transfer between screens - Exportable image and PDF formats Sources: - Uizard pricing page, https://uizard.io/pricing, verified 2026-08-10 FAQ: - Q: Can Uizard build a working app on its own? A: No, it produces wireframes and design assets, not a running application. Voters use it before handing the brief to a builder like Base44 or a developer. Opinions (2): - Nadia Okonkwo (@nadiao) [Editor panel], score 680/1000, 23 found useful, 2026-08-09: We use it at the very start of a project to get stakeholders aligned on a layout before anyone writes a brief for a real build. It has never shipped anything for us directly and it has never claimed it would. - Agent Verdict Panel (@agentverdict) [Editor panel], score 660/1000, 18 found useful, 2026-08-06: One of the longer running tools on this board and the input handling shows it: a rough sketch or a screenshot both turned into an editable wireframe without complaint. There is no production code export, so it is scored here as an ideation tool rather than a builder, and it does that specific job well. --- ## Zite URL: https://www.agent-verdict.com/tool/zite Opinions page: https://www.agent-verdict.com/tool/zite/opinions Official site: https://zite.com Category: AI app builder Tier: B Community score: 687 of 1000 Votes: 36 total, 25 up, 11 down Board rank: 27 Trend this week: down 19 Head to head record: no duels yet Price from: USD 15 / month Last verified: 2026-08-22 Zite is the gentlest entry point on the board. The onboarding assumes nothing, the editor is legible to someone who has never seen a component tree, and getting a marketing site or a light data driven app live takes minutes. That accessibility is why it climbed a place this week on value and design votes. It is also the ceiling: builders who need custom logic, real relations or an owned codebase run out of room quickly and vote accordingly. Pricing: Apps and users are unlimited on every tier, including the free one, so the paid ladder does not buy headroom on either. Pro at USD 15 a month raises monthly credits, database records and workflow runs, adds one custom domain and removes Zite branding. Business at USD 55 lifts those quotas again and adds unlimited custom domains and white label emails. Team and Enterprise are named above Business without a published price. Strengths: - Easiest onboarding on the board by a wide margin - Low paid entry point at USD 15 per month - Fast for marketing sites and light data apps - Climbing on value and design votes Weaknesses: - Low ceiling for custom logic - Data modelling is shallow - Limited code ownership story Features: - Prompt to site and light app generation - Visual editor with no code concepts - Forms and simple data collections - Custom domains - Template library Sources: - Zite official site, https://zite.com, verified 2026-08-12 FAQ: - Q: Is Zite an app builder or a site builder? A: Mostly a site builder with light app capability. On this board it competes on value and accessibility rather than depth. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 690/1000, 20 found useful, 2026-08-12: We handed this to a non technical tester with no instructions and they had a live site in nine minutes, which nothing else on the board managed. Then we asked for a related data model and hit the ceiling immediately. Priced and scoped honestly, and B tier is the right place for it. - Gemma Ruiz (@gemmar) [Editor panel], score 720/1000, 15 found useful, 2026-08-15: Twelve euros a month and my booking page has been up for six weeks without me touching it. I do not need a database, I need a page that works. It is the cheapest thing here that does not feel cheap. --- ## Softgen URL: https://www.agent-verdict.com/tool/softgen Opinions page: https://www.agent-verdict.com/tool/softgen/opinions Official site: https://softgen.ai Category: AI app builder Tier: B Community score: 681 of 1000 Votes: 22 total, 15 up, 7 down Board rank: 28 Trend this week: no change Head to head record: no duels yet Price from: USD 25 / month Last verified: 2026-08-22 Softgen follows the same chat to full stack app pattern as the category leaders on a smaller team and a smaller community. Straightforward CRUD style briefs come out working and reasonably quickly. Testers reported the agent needing more manual correction on multi step briefs than the top of the board, and the template and integration gallery is noticeably thinner, which is the main thing separating it from the leaders on the reliability and integrations axes. Pricing: The pricing page publishes no free plan, only a free signup flow. Starter is USD 25 a month for 500 credits across 10 projects, and Pro is USD 49 a month for 1,500 credits and unlimited projects. Strengths: - Handles straightforward CRUD style briefs reliably - Simple, familiar chat to app workflow - Accessible pricing for early stage use Weaknesses: - More manual correction needed on multi step briefs - Thinner template and integration gallery than the leaders - Smaller community, fewer answers to common problems Features: - Prompt to full stack web application - Built in database and basic authentication - Conversational iteration - One click deploy Sources: - Softgen official site, https://softgen.ai, verified 2026-08-08 FAQ: - Q: How does Softgen compare to Lovable or Bolt on a hard brief? A: It handles simple briefs comparably well but needs more manual correction as complexity rises, which is reflected in its lower reliability score against the category leaders. Opinions (2): - Petros Ioannou (@pioannou) [Editor panel], score 705/1000, 12 found useful, 2026-08-08: Cheap enough that I did not think twice about trying it for an early prototype, and it got me something working. When I hit a snag there were fewer forum threads and fewer people to ask than with the bigger names, which is the real cost of using something smaller. - Agent Verdict Panel (@agentverdict) [Editor panel], score 650/1000, 11 found useful, 2026-08-04: Same chat to full stack pattern as the category leaders on a visibly smaller team and community. Straightforward CRUD briefs came out working. Multi step briefs needed more manual correction than the top of the board, and the template and integration gallery is noticeably thinner, which is exactly what shows up in its reliability and integrations scores. --- ## Floot URL: https://www.agent-verdict.com/tool/floot Opinions page: https://www.agent-verdict.com/tool/floot/opinions Official site: https://floot.com Category: AI app builder Tier: B Community score: 674 of 1000 Votes: 20 total, 14 up, 6 down Board rank: 29 Trend this week: no change Head to head record: no duels yet Price from: USD 25 / month Last verified: 2026-08-22 Floot's pitch is that it treats UI polish and backend logic as one job rather than generating a nice screen with placeholder data behind it. On simple apps that shows in genuinely usable forms and data flows out of the first prompt. It is early: the community is small, the template gallery is thin, and voters report more trial and error getting a multi screen application to hold together than with the established leaders. Pricing: Free tier for early testing. Paid plans start around USD 25 per month for more generations and published projects. Strengths: - UI polish and working backend logic treated as one job - Genuinely usable data flows out of a simple first prompt - Actively developed, features shipping frequently Weaknesses: - Small community and thin template gallery given how new it is - More trial and error on multi screen applications - Less battle tested than the established leaders Features: - Prompt to full stack app with working data flows - Built in database and forms - Conversational iteration - One click publish Sources: - Floot official site, https://floot.com, verified 2026-08-07 FAQ: - Q: Is Floot mature enough for a real project? A: For a simple app, testers found it usable out of the gate. For anything with several connected screens, expect more iteration than with a more established builder. Opinions (2): - Ravi Chandrasekaran (@ravic) [Editor panel], score 700/1000, 11 found useful, 2026-08-07: Feels like a tool that is actively getting better every time I open it, and my simple app's data flows worked first try, which surprised me for something this new. I would not bet a client project on it yet, mainly because there is nobody else's experience to check mine against. - Agent Verdict Panel (@agentverdict) [Editor panel], score 645/1000, 9 found useful, 2026-08-03: Floot's pitch is treating UI polish and backend logic as one job rather than a nice screen over placeholder data, and on a simple app that showed in genuinely usable forms straight out of the first prompt. It is early: small community, thin template gallery, and a multi screen application took more trial and error to hold together than with an established leader. --- ## Durable URL: https://www.agent-verdict.com/tool/durable Opinions page: https://www.agent-verdict.com/tool/durable/opinions Official site: https://durable.com Category: AI app builder Tier: B Community score: 654 of 1000 Votes: 38 total, 25 up, 13 down Board rank: 30 Trend this week: no change Head to head record: no duels yet Price from: USD 25 / month Last verified: 2026-08-22 Durable is built for the smallest, fastest case: a service business that needs a site today. One prompt produces pages, written copy and a logo together, and a built in CRM and invoicing layer means some owners never need a second tool. Voters mark it down hard on scope and design ceiling, since it is not attempting a custom application and the visual output is templated rather than distinctive. Pricing: A CRM ships on every plan, the free one included, where it holds up to ten customers on a durable.site address alongside 5 images and 10 AI messages a month. Launch at USD 25 a month is where the customer list becomes unlimited and a custom domain is included, free up to USD 20 of value, together with automatic lead capture and replies. Grow at USD 49 adds unlimited users and ten times the AI usage. On annual billing those two become USD 22 and USD 41 a month. Strengths: - Fastest path from nothing to a published small business site - Copy and logo generated alongside the pages - Built in CRM and invoicing, genuinely useful for a service business Weaknesses: - Design output is templated, not distinctive - No path to a custom application beyond a small business site - Limited customisation compared to a real visual editor Features: - Prompt to full site with copy and logo - Built in CRM and invoicing - Booking and payment collection for service businesses - One click publish with a custom domain Sources: - Durable pricing page, https://durable.com/pricing, verified 2026-08-09 FAQ: - Q: Is Durable meant for anything beyond a small business website? A: No. It is optimised for the fastest possible small business site with copy, logo and basic business tools included, and it is honest about not attempting a custom application. Opinions (2): - Petra Novak (@petranovak) [Editor panel], score 690/1000, 21 found useful, 2026-08-08: Running events for a living, I needed a site and a way to invoice clients, not a development project. Durable gave me both in the same afternoon and I have not opened another tool since. I know it is not making anyone a custom app and I do not need it to. - Agent Verdict Panel (@agentverdict) [Editor panel], score 630/1000, 17 found useful, 2026-08-05: The fastest path from nothing to a published small business site we tested, copy and logo generated together, and the built in CRM and invoicing genuinely replaced a second tool for the panel's mock landscaping business. It is not attempting a custom application at all, and the visual output is templated rather than distinctive, which is what keeps it out of A tier. --- ## Hostinger Horizons URL: https://www.agent-verdict.com/tool/hostinger-horizons Opinions page: https://www.agent-verdict.com/tool/hostinger-horizons/opinions Official site: https://hostinger.com Category: AI app builder Tier: C Community score: 633 of 1000 Votes: 45 total, 29 up, 16 down Board rank: 31 Trend this week: no change Head to head record: no duels yet Price from: USD 10.99 / month at renewal Last verified: 2026-08-22 Hostinger Horizons lets existing Hostinger customers generate a simple web app from a prompt without leaving the hosting dashboard they already use, which is the entire value proposition. Generation quality is serviceable for a basic app or landing tool but clearly a bundled feature rather than the product's main focus, and voters comparing it against a dedicated app builder consistently rate the standalone tools higher on every axis except convenience for an existing customer. Pricing: Included with qualifying Hostinger hosting plans starting around USD 10.99 per month at renewal, rather than sold as a standalone product. Strengths: - Convenient if you already host with Hostinger - No separate account or billing relationship needed - Fine for a basic app or landing page Weaknesses: - Clearly a bundled feature, not the product's main focus - Generation quality trails the dedicated app builders - Limited value if you are not already a Hostinger customer Features: - Prompt to simple web app inside the hosting dashboard - Bundled with existing Hostinger hosting plans - Basic database and form support - One click deploy to Hostinger hosting Sources: - Hostinger Horizons page, https://hostinger.com, verified 2026-08-08 FAQ: - Q: Should I use Horizons if I am not a Hostinger customer? A: Voters say no. It is a convenience feature for existing customers, and a dedicated app builder rates higher on this board for anyone choosing a tool from scratch. Opinions (2): - Corinne Faucher (@cfaucher) [Editor panel], score 655/1000, 25 found useful, 2026-08-07: I already host four sites there so trying the app builder cost me nothing extra, and it made a basic booking form I needed without a second bill. I would not have chosen it on its own merits if I were shopping from scratch. - Agent Verdict Panel (@agentverdict) [Editor panel], score 610/1000, 21 found useful, 2026-08-04: Convenient exactly once: if you are already paying for Hostinger hosting, generating a simple app without leaving the dashboard is genuinely useful. Compared against a dedicated app builder on the same brief it trailed on every axis except that one convenience, which is the honest read of a bundled feature rather than a product built around the prompt. --- ## 10Web URL: https://www.agent-verdict.com/tool/10web Opinions page: https://www.agent-verdict.com/tool/10web/opinions Official site: https://10web.io Category: AI app builder Tier: C Community score: 626 of 1000 Votes: 32 total, 20 up, 12 down Board rank: 32 Trend this week: no change Head to head record: no duels yet Price from: USD 10 / website / month, billed annually Last verified: 2026-08-22 10Web's generator produces a working WordPress site from a prompt, which matters to anyone who wants the world's most common CMS underneath rather than a locked platform. Voters who already know WordPress like being able to drop into the plugin ecosystem afterward. It scores low on this board because a generated WordPress site inherits WordPress's own maintenance and performance overhead, and the AI generation step itself is behind the newer builders on speed and design taste. Pricing: Paid plans start around USD 10 per website per month, billed annually, including managed WordPress hosting, with higher tiers adding more sites and performance optimisation. Strengths: - Outputs a real WordPress site, not a proprietary format - Access to the full WordPress plugin ecosystem afterward - Familiar territory for anyone who already manages WordPress Weaknesses: - Inherits WordPress's own maintenance and performance overhead - Generation speed and design taste behind the newer builders - No meaningful free tier to try before committing Features: - Prompt to generated WordPress site - Managed WordPress hosting included - AI performance and image optimisation - Standard WordPress plugin compatibility Sources: - 10Web pricing page, https://10web.io/pricing, verified 2026-08-07 FAQ: - Q: Why choose 10Web over a purpose built AI app builder? A: Mainly if you specifically want a WordPress site for the plugin ecosystem or existing WordPress skills. For a new project with no WordPress requirement, voters rate the newer builders higher. Opinions (2): - Trevor Higgs (@thiggs) [Editor panel], score 650/1000, 18 found useful, 2026-08-06: For a client who insists on WordPress, this is a faster starting point than building a theme from scratch by hand. I still spend the same afternoon I always spend on WordPress performance tuning afterward, the AI did not remove that step. - Agent Verdict Panel (@agentverdict) [Editor panel], score 600/1000, 15 found useful, 2026-08-03: A real, working WordPress site out of a prompt, which matters specifically if you want the plugin ecosystem underneath rather than a locked platform. Generation speed and design taste both trailed the newer builders, and a generated WordPress site still inherits WordPress's own maintenance overhead, which voters clearly weigh against it. --- ## B12 URL: https://www.agent-verdict.com/tool/b12 Opinions page: https://www.agent-verdict.com/tool/b12/opinions Official site: https://b12.io Category: AI app builder Tier: C Community score: 618 of 1000 Votes: 26 total, 16 up, 10 down Board rank: 33 Trend this week: no change Head to head record: no duels yet Price from: USD 49 / month Last verified: 2026-08-22 B12 draws a first draft site with AI and then routes it through a human review step before handing it back, aimed at consultants, coaches and other service businesses who want a polished result without doing the work themselves. That human step raises the design floor compared to a pure AI generator, but it also means it is slower to first result than every other tool on this board and it is not attempting to be a general application builder. Pricing: There is a permanent free plan, not just a trial: it publishes to a b12sites.com subdomain on 5 AI credits a day, capped at 30 a month. Paid plans start around USD 49 per month, with pricing reflecting the human review built into the workflow. Strengths: - Human review step raises the design floor above a pure AI draft - Good fit for consultants and service businesses without design skill - Bookings, contracts and invoicing included for that audience Weaknesses: - Slowest time to first result of anything on this board - Not attempting to be a general purpose application builder - Ongoing changes route through the same slower review flow Features: - AI draft plus human review before delivery - Built in bookings, contracts and invoicing - Service business focused templates - Ongoing support for edits after launch Sources: - B12 pricing page, https://www.b12.io/pricing/, verified 2026-08-24 FAQ: - Q: How fast is a B12 site ready? A: Slower than a pure AI generator because a human reviews the draft before you see it. Voters trade that speed for a more polished first result. Opinions (2): - Simone Laurent (@slaurent) [Editor panel], score 645/1000, 14 found useful, 2026-08-05: I am not a designer and I did not want to learn to be one, so paying for a person to check the AI's work before I see it is worth it to me. Everyone should know going in that ongoing edits go through that same slower review, it is not a one time wait. - Agent Verdict Panel (@agentverdict) [Editor panel], score 585/1000, 12 found useful, 2026-08-02: The human review step is the whole story: a real person checks the AI draft before you see it, which raises the design floor above a pure generator and also makes it the slowest time to first result of anything we tested on this board, by a wide margin. It is not attempting to be a general application builder and it does not pretend otherwise. --- ## Typedream URL: https://www.agent-verdict.com/tool/typedream Opinions page: https://www.agent-verdict.com/tool/typedream/opinions Official site: https://typedream.com Category: AI app builder Tier: C Community score: 611 of 1000 Votes: 23 total, 14 up, 9 down Board rank: 34 Trend this week: no change Head to head record: no duels yet Price from: USD 20 / site / month, billed monthly Last verified: 2026-08-22 Typedream builds pages from stacked content blocks in a style close to Notion, with an AI layer that drafts copy and page structure on request. It is quick and cheap for a simple marketing site, landing page or personal site. Voters mark it well down on this board's brief because there is no application logic or data model at all, only content pages, which caps its relevance for anything beyond a simple site. Pricing: Publishing on the free plan is capped at a single page on a typedream.app subdomain. Launch costs USD 20 per site per month billed monthly, or USD 15 billed annually, and lifts the page cap entirely while adding a custom domain, SEO and code injection, a CMS blog, and cutting the transaction fee from 5 percent to 2 percent. Strengths: - Quick and cheap for a simple marketing or personal site - Familiar block based editing if you already use Notion - AI assisted copy drafting speeds up the blank page Weaknesses: - No application logic or real data model - Design ceiling is low compared to a dedicated site builder - Not attempting anything close to the board's application brief Features: - Block based page building with AI drafting - Simple forms for basic data capture - Custom domain publishing - SEO basics built in Sources: - Typedream pricing page, https://typedream.com/pricing, verified 2026-08-05 FAQ: - Q: Can Typedream build an application? A: No, it builds content pages only. It appears on this board because enough builders compare it for the simplest end of the brief, a personal or marketing site. Opinions (2): - Amara Nwosu (@anwosu) [Editor panel], score 630/1000, 13 found useful, 2026-08-04: My personal site and a small portfolio both live here and it took an evening, not a weekend. I use Notion for everything else already so the block editor felt familiar on day one. - Agent Verdict Panel (@agentverdict) [Editor panel], score 580/1000, 11 found useful, 2026-08-01: Quick and cheap for exactly what it is, a Notion style block based page builder with an AI layer drafting copy on request. There is no application logic or data model at all, only content pages, which caps how much of this board's brief it can even attempt and is why it sits near the bottom despite being pleasant to use. --- ## Wegic URL: https://www.agent-verdict.com/tool/wegic Opinions page: https://www.agent-verdict.com/tool/wegic/opinions Official site: https://wegic.ai Category: AI app builder Tier: C Community score: 604 of 1000 Votes: 19 total, 11 up, 8 down Board rank: 35 Trend this week: no change Head to head record: no duels yet Price from: USD 23.90 / month, billed annually Last verified: 2026-08-22 Wegic builds and then continues editing a website entirely through chat, rather than handing you a visual canvas once the first draft lands. That keeps the workflow simple for someone who does not want to learn an editor, and testers found it fine for a basic multi page site. It has no real application layer and the conversational only editing became a limitation once changes got specific, since some layout adjustments are easier to describe than to get right through text alone. Pricing: Signup credits let you chat with the AI, but the vendor states that generating a site is a paid feature, so there is no usable free tier. Starter is USD 23.90 a month billed annually and keeps the site on a permanent .wegic.net subdomain; a custom domain only arrives on the Premium plan above it. Strengths: - Simple chat only workflow with no editor to learn - Fine for a basic multi page marketing site - Continuous editing through further conversation Weaknesses: - No visual canvas, some changes are hard to describe accurately in text - No real application layer or data model - Smaller community than the established site builders Features: - Chat to multi page website - Ongoing edits through conversation - Basic contact forms - One click publish with a custom domain Sources: - Wegic official site, https://wegic.ai, verified 2026-08-04 FAQ: - Q: Does Wegic have a visual editor? A: No, all editing happens through chat. Voters who want direct visual control point to Framer or Webflow instead. Opinions (2): - Julio Restrepo (@jrestrepo) [Editor panel], score 615/1000, 10 found useful, 2026-08-03: For a client who just wants a few pages and does not care about pixel perfect control, chatting the whole site into existence is genuinely fast. I hit the wall the first time a client wanted one button moved three pixels and there was no way to just drag it. - Agent Verdict Panel (@agentverdict) [Editor panel], score 570/1000, 9 found useful, 2026-07-31: Chat only, no visual canvas at all, which keeps the workflow genuinely simple for someone who does not want to learn an editor and becomes a real limitation the moment a change is easier to point at than to describe. Fine for a basic multi page site, nothing close to an application layer. --- ## Mixo URL: https://www.agent-verdict.com/tool/mixo Opinions page: https://www.agent-verdict.com/tool/mixo/opinions Official site: https://mixo.io Category: AI app builder Tier: C Community score: 590 of 1000 Votes: 21 total, 12 up, 9 down Board rank: 36 Trend this week: no change Head to head record: no duels yet Price from: USD 20 / month Last verified: 2026-08-22 Mixo turns a plain description of a business into a published marketing site with pages, a contact form, a blog and search optimisation handled for you, and the vendor aims it squarely at local trades and professional services rather than at software launches. There is no application layer at all: no auth, no data model, no server side logic and no code to take away, so voters treat it as a lead capture site builder rather than anything competing with the app builders on this board's brief. Pricing: The free plan stays live indefinitely on a Mixo branded address, carrying the Mixo badge, a live site and a contact form. Business at USD 20 per month is the entry paid plan: it buys a free custom domain, removes the badge, and unlocks multi page sites and enquiry management. Premium at USD 40 adds deeper site analytics, further lead capture tools, brand matching, team access and tool integrations. Strengths: - Published site with pages, a contact form and a blog in about a minute - Search and AI search optimisation shipped by default, including an automatic llms.txt - Cheap entry price Weaknesses: - No application layer at all: no auth, no data model and no server side logic - Design templates feel dated next to the newer builders - No code export and no GitHub, so nothing can be moved off Mixo hosting Features: - Prompt to a multi page business site with copy - Contact forms with enquiry tracking and contact export - Basic analytics on page visits - One click publish with a custom domain Sources: - Mixo official site, https://mixo.io, verified 2026-08-02 FAQ: - Q: Is Mixo still relevant next to newer AI builders? A: For a small business that needs a findable site with a contact form, yes, it still does that job quickly and cheaply. For anything with a real data model, or code you can take away, voters point to almost anything else on this board. Opinions (2): - Colin Aspen (@caspen) [Editor panel], score 600/1000, 12 found useful, 2026-08-01: Used it to stand up a real page for an idea before writing a single line of the actual product, contact form and all, in under ten minutes. I would never mistake it for anything more than that and it has never tried to be. - Agent Verdict Panel (@agentverdict) [Editor panel], score 555/1000, 9 found useful, 2026-07-29: A name and a one line description into a published marketing site with a contact form, and it still does that narrow job quickly. There is no application layer of any kind, no auth and no data model, and the templates read a little dated next to what the newer builders default to. --- ## Trickle URL: https://www.agent-verdict.com/tool/trickle Opinions page: https://www.agent-verdict.com/tool/trickle/opinions Official site: https://trickle.so Category: AI app builder Tier: C Community score: 587 of 1000 Votes: 27 total, 16 up, 11 down Board rank: 37 Trend this week: down 29 Head to head record: no duels yet Price from: USD 17 / month Last verified: 2026-08-22 Trickle is scoped deliberately small and the board scores it accordingly. Landing pages, forms and light workflows come out quickly and cheaply, and for that narrow job the votes are positive. Asked to behave like a full application builder it falls over, which is why it anchors C tier rather than competing in B. It is on the board because a lot of people genuinely start here, and because a scoreboard that only lists winners is not a scoreboard. Pricing: Free tier for basic pages. Paid plans start around USD 17 per month, with limits on projects and AI generations. Strengths: - Low paid entry point at USD 17 per month - Very fast for pages, forms and small workflows - Almost no learning curve Weaknesses: - Not a real application builder - No meaningful data modelling - Weakest head to head record on the board Features: - AI page and form generation - Simple workflow automations - Template gallery - Custom domain on paid plans Sources: - Trickle official site, https://trickle.so, verified 2026-08-11 FAQ: - Q: Why is Trickle in C tier? A: C tier is not a failing grade, it is a scope statement. Trickle does pages, forms and small workflows well and does not attempt the application work that S and A tier entries are voted on. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 590/1000, 15 found useful, 2026-08-11: Judged as a page and form builder it is quick, cheap and fine. Judged against the brief every other tool on this board completed it does not finish, which is what C tier records. No complaints about the product, it simply is not attempting the same job. - Ines Duarte (@inesd) [Editor panel], score 610/1000, 11 found useful, 2026-08-15: I use it for campaign landing pages and lead forms and it does that in minutes. I tried to make it into a client portal and that was my mistake, not the tool's. --- ## Onlook URL: https://www.agent-verdict.com/tool/onlook Opinions page: https://www.agent-verdict.com/tool/onlook/opinions Official site: https://onlook.com Category: AI app builder Tier: C Community score: 578 of 1000 Votes: 18 total, 10 up, 8 down Board rank: 38 Trend this week: no change Head to head record: no duels yet Price from: Not published by the vendor Last verified: 2026-08-22 Onlook is not a generator, it is a visual editor for a codebase you already have: point it at a React project and you can move, restyle and rewire real components on a live canvas instead of hand editing files. Voters who tried it as a full app builder found the ceiling low, since it assumes an existing project rather than starting one, which is a different job than most tools on this board are being judged for. Pricing: The vendor page states Onlook is currently in closed beta, with a demo booking and custom team pricing only. A closed beta is not a free tier. Checked August 21, 2026. Strengths: - Visual editing of a live, real React codebase - Open source, no lock in to a hosted platform - Useful alongside a builder rather than instead of one Weaknesses: - Assumes an existing project, does not start one from a prompt - Early stage, rougher edges than a mature product - Not a fit for anyone wanting a full build from nothing Features: - Visual canvas editing of live React components - Open source and self hostable - Works alongside an existing codebase - Direct code changes from visual edits Sources: - Onlook official site, https://onlook.com, verified 2026-08-01 FAQ: - Q: Can Onlook build an app from a prompt like the other tools here? A: No, it edits an existing React codebase visually rather than generating a new one. Voters use it after a build from another tool, not instead of one. Opinions (2): - Femi Adeyemi (@femia) [Editor panel], score 610/1000, 10 found useful, 2026-07-31: I use it alongside a builder from higher up this board, generate the scaffold elsewhere and then use this to actually feel out the visual details on the live app. Rougher around the edges than a mature product, which is fair for something this early. - Agent Verdict Panel (@agentverdict) [Editor panel], score 545/1000, 8 found useful, 2026-07-28: Not a generator at all, a visual editor for a React codebase you already have, open source and genuinely useful for restyling live components without hand editing files. Judged as a full app builder the ceiling is low because it assumes an existing project rather than starting one, which is a different job than most of this board is being scored on. --- ## Visily URL: https://www.agent-verdict.com/tool/visily Opinions page: https://www.agent-verdict.com/tool/visily/opinions Official site: https://visily.ai Category: AI app builder Tier: C Community score: 573 of 1000 Votes: 16 total, 9 up, 7 down Board rank: 39 Trend this week: no change Head to head record: no duels yet Price from: USD 11 / editor / month, billed annually Last verified: 2026-08-22 Visily takes a screenshot of an app you like or a short prompt and produces an editable wireframe, used mostly by product teams to sketch an idea quickly before committing to a real build. It does export code on Pro and above, but there is no working application behind the wireframes, which is why it sits at the lower end of a board scored on shipping a working product rather than a design step. Pricing: Starter is USD 0 and the cap is on boards, not projects: two editable boards per workspace, 2,500 elements per board and 300 AI credits a month, with image exports limited to 1x and watermarked. Pro is USD 11 per editor a month billed annually for unlimited boards, 3,000 credits per editor, Figma import and export, code export, the Visily MCP server and 7 day version history, which Starter has none of. Commenting and real time co-designing are on every tier including the free one; what Pro adds is inviting guest viewers and editors from outside the workspace. Strengths: - Fast from a screenshot or prompt to an editable wireframe - Useful for product teams sketching ideas before a build - Simple enough for non designers to use directly Weaknesses: - Code export is gated to Pro and above, and the vendor does not disclose the output format - No working application behind the wireframes - Does not attempt the board's working application brief Features: - Screenshot or prompt to wireframe - Editable design canvas - Team collaboration on shared boards - Exportable design assets Sources: - Visily pricing page, https://visily.ai/pricing, verified 2026-07-31 FAQ: - Q: Should I use Visily instead of an app builder? A: No, it is a design ideation step. Voters use it to sketch an idea, then hand the result to a builder or a developer to actually ship. Opinions (2): - Bianca Ferrante (@bferrante) [Editor panel], score 590/1000, 9 found useful, 2026-07-30: Our team uses this to get everyone looking at the same rough idea before a single ticket gets written. It has never been anywhere near production and nobody on my team has ever expected it to be. - Agent Verdict Panel (@agentverdict) [Editor panel], score 540/1000, 7 found useful, 2026-07-27: A screenshot or a short prompt turns into an editable wireframe quickly, and it is genuinely useful for a product team sketching an idea before committing to a real build. Code export exists on Pro and above, but there is no working application behind the wireframes, which is the entire reason it sits at the lower end of a board scored on shipping. --- ## CatDoes URL: https://www.agent-verdict.com/tool/catdoes Opinions page: https://www.agent-verdict.com/tool/catdoes/opinions Official site: https://catdoes.com Category: AI app builder Tier: C Community score: 560 of 1000 Votes: 13 total, 7 up, 6 down Board rank: 40 Trend this week: no change Head to head record: no duels yet Price from: USD 17 / month Last verified: 2026-08-22 CatDoes is one of the smaller entries on this board, a prompt driven builder for simple sites and micro apps with a following that is real but narrow. Testers got a working basic result quickly on a small brief. It has the smallest template gallery and community of any tool tracked here, and voters could not get a multi step brief past the first draft without heavy manual cleanup. Pricing: The free plan allows one public project, deployed as a web app, on 25 credits. Core at USD 17 per month is the entry paid plan and buys 50 monthly credits, a second project, private projects, the file browser and GitHub import. The capabilities the product leads with cost considerably more: a custom domain, App Store and Google Play releases and removal of the CatDoes badge are all listed from Starter at USD 42 per month, and none of them appears in Core's own feature list. Strengths: - Quick working result on a small, simple brief - Growing community despite its size - Cheap entry price Weaknesses: - Smallest template gallery and community on the board - Multi step briefs need heavy manual cleanup - Limited track record compared to established tools Features: - Prompt to simple site or micro app - Basic data storage for small projects - One click publish - Community template sharing Sources: - CatDoes official site, https://catdoes.com, verified 2026-07-30 FAQ: - Q: Is CatDoes ready for a serious project? A: For a small, simple brief, testers got a usable first result. For anything bigger, expect to do more of the work manually than with an established builder. Opinions (2): - Toby Ellison (@tellison) [Editor panel], score 585/1000, 7 found useful, 2026-07-29: Found it through a small community that has grown around it and it did exactly what I needed for a weekend project. I would not describe the ecosystem around it as mature yet, there just are not many other people to compare notes with. - Agent Verdict Panel (@agentverdict) [Editor panel], score 520/1000, 6 found useful, 2026-07-26: One of the smaller entries here, a working basic result on a small simple brief came together quickly. The smallest template gallery and community of anything we tracked, and a multi step brief did not survive past the first draft without heavy manual cleanup on our end. --- ## YouWare URL: https://www.agent-verdict.com/tool/youware Opinions page: https://www.agent-verdict.com/tool/youware/opinions Official site: https://youware.com Category: AI app builder Tier: C Community score: 546 of 1000 Votes: 12 total, 7 up, 5 down Board rank: 41 Trend this week: no change Head to head record: no duels yet Price from: Not published by the vendor Last verified: 2026-08-22 YouWare is closer to a community gallery than a product builder: prompts generate small interactive sites, games and mini apps that are then shared and remixed by other users on the platform. It is genuinely fun for quick, disposable creations, and voters treat it that way. It scores at the bottom of this board because it is not aimed at producing a serious, standalone application, and there is no real path to owning and deploying the result independently. Pricing: youware.com/pricing returns HTTP 200 with a rendered 404 body. No entry price and no free tier statement is published by the vendor. Checked August 21, 2026. Strengths: - Fun, fast way to generate small interactive creations - Community remixing makes iteration social rather than solitary - No cost barrier to trying it Weaknesses: - Not aimed at a serious, standalone application - No real path to owning and deploying the result independently - Lowest scoring reliability and integrations on the board Features: - Prompt to shareable mini app, game or site - Community remixing of published creations - Built in gallery and discovery feed - No account needed to browse Sources: - YouWare official site, https://youware.com, verified 2026-07-29 FAQ: - Q: Why does YouWare sit at the bottom of the board? A: It is built for sharing quick, disposable creations within a community, not for shipping and owning a serious application, which is what this board's weights reward. Opinions (2): - Iris Solheim (@isolheim) [Editor panel], score 570/1000, 7 found useful, 2026-07-28: I have made a handful of silly little games here just to see what other people remix them into, and that part is genuinely fun. I would never point at this if someone asked me how to build a real product. - Agent Verdict Panel (@agentverdict) [Editor panel], score 500/1000, 5 found useful, 2026-07-25: Closer to a community gallery than a product builder, prompts generate small interactive sites and games that get shared and remixed by other users on the platform. Genuinely fun for a disposable creation, and there is no real path to owning and deploying the result independently, which is the entire reason it anchors the bottom of this board. --- ## Google Stitch URL: https://www.agent-verdict.com/tool/google-stitch Opinions page: https://www.agent-verdict.com/tool/google-stitch/opinions Official site: https://stitch.withgoogle.com Category: AI app builder Tier: C Community score: 535 of 1000 Votes: 29 total, 16 up, 13 down Board rank: 42 Trend this week: no change Head to head record: no duels yet Price from: No cost Last verified: 2026-08-22 Google Stitch generates UI designs and basic frontend code from a text prompt or a rough sketch, and it carries the label a lot of early Labs experiments do: interesting direction, rough execution. Testers found the generated designs a reasonable starting point for ideation, but the exported code needed substantial rework before it resembled anything production ready, which is what keeps it at the bottom of the board on reliability and code quality. Pricing: Free to use as a Google Labs experiment, with no committed pricing plan since it has not left beta. Strengths: - Free to try, backed by Google's design and model research - Reasonable starting point for early UI ideation - Accepts a rough sketch as well as a text prompt Weaknesses: - Exported code needs substantial rework before production use - Still clearly a beta, features change without notice - No committed pricing or roadmap since it has not left Labs Features: - Prompt or sketch to UI design - Basic frontend code export - Experimental Google Labs feature set - Iteration through further prompts Sources: - Google Stitch by Labs, https://stitch.withgoogle.com, verified 2026-07-28 FAQ: - Q: Is Google Stitch ready for production use? A: Not yet, by Google's own framing. Testers treat it as an ideation tool and expect to substantially rework any exported code before shipping it. Opinions (2): - Agent Verdict Panel (@agentverdict) [Editor panel], score 490/1000, 16 found useful, 2026-07-24: A Google Labs experiment and it carries the label honestly: a rough sketch or a prompt produced a reasonable starting point for UI ideation, and the exported frontend code needed substantial rework before it resembled anything production ready. Still clearly a beta, features changed under us mid test, which is the biggest reason it sits at the bottom of the board on reliability. - Grace Petit (@gracep) [Editor panel], score 555/1000, 12 found useful, 2026-07-27: Free, backed by real design research, and worth trying for the earliest stage of an idea. I rewrote most of the exported code by hand before shipping anything, which Google's own framing about it being a Labs experiment prepares you for if you read it. --- # Head to head duels ## lovable against bolt-new URL: https://www.agent-verdict.com/head-to-head/lovable/bolt-new Vote split: 43 to 38 Winner: lovable Published: 2026-08-18 Design taste against raw start speed Bolt.new wins the stopwatch. WebContainers mean there is no remote sandbox to warm up, and on repeated cold starts it is the fastest thing on the board to a running dev server. Lovable wins the render. Voters describe the difference as the gap between a page you would show a client and a page you would show a teammate. Over 81 votes that gap is worth more than the seconds Bolt saves, 53.1 percent to 46.9 percent. Where each one flips the result Bolt takes the matchup for anyone iterating dozens of times an hour, because reload cost compounds. Lovable takes it for anyone who has to present the output. Both export to GitHub, so neither result is a lock in decision. ## claude-code against cursor URL: https://www.agent-verdict.com/head-to-head/claude-code/cursor Vote split: 41 to 38 Winner: claude-code Published: 2026-08-18 The coding agent final 51.9 percent to 48.1 percent over 79 votes, the tightest result in the category. Claude Code takes the long task. Voters consistently report better retention across a refactor that spans an afternoon and better behaviour when it has to read before it writes. Cursor takes the minute to minute. Predictive multi line completion is still the best in the category and the editor asks nothing of people who do not live in a terminal. Neither result is a recommendation to pick one. The most common pattern in the comments is Cursor for editing and Claude Code for anything that needs a plan. ## framer against webflow URL: https://www.agent-verdict.com/head-to-head/framer/webflow Vote split: 31 to 27 Winner: framer Published: 2026-08-17 Two design led site builders, one built around the prompt from day one Both ship real semantic markup rather than a page builder's proprietary soup, which is why voters put them head to head instead of against the app builders. The 53.4 to 46.6 split comes down to which one the AI layer was built for. Framer's AI mode was designed around the prompt from the start, and it shows in how little cleanup the first draft needs. Webflow's generator is a strong addition to a decade old visual canvas, and it still wins on CMS depth for anyone running a large content operation. Where each one wins Pick Framer for a fast, distinctive marketing site or portfolio. Pick Webflow if the project is content heavy enough that CMS collections, localisation and enterprise publishing controls matter more than how quickly the first draft looks right. ## totalum against base44 URL: https://www.agent-verdict.com/head-to-head/totalum/base44 Vote split: 24 to 15 Winner: totalum Published: 2026-08-17 Two database first builders, one clear split Both of these arrive with the backend already connected, which is why they attract the same voters. The 61.5 to 38.5 split comes down to what happens on day ten. Totalum takes reliability and code ownership. The schema is a real relational model, changes propagate, and the whole project exports as Next.js source you can host yourself. Base44 takes value and setup speed. Nothing to configure, nothing to connect, and a cheaper entry point. The complaint that decides the duel is the exit: when the opinionated stack does not anticipate your requirement, the workaround is harder than it would be on an owned codebase. ## bolt-new against v0 URL: https://www.agent-verdict.com/head-to-head/bolt-new/v0 Vote split: 39 to 35 Winner: bolt-new Published: 2026-08-16 Breadth against polish v0 produces the code a senior reviewer would approve without comment. Bolt.new produces a whole running application, backend included, in the time v0 takes to nail the interface. The 52.7 to 47.3 split is the closest duel on the board. Voters building a product pick Bolt. Voters building the hard screen inside an existing product pick v0, and several said in comments that they use both in the same week. ## rork against a0-dev URL: https://www.agent-verdict.com/head-to-head/rork/a0-dev Vote split: 14 to 12 Winner: rork Published: 2026-08-15 The mobile only duel Both generate a real React Native app through Expo rather than a web page in a shell, which is the entire reason either of them is worth comparing instead of pointing everyone at a web app builder. The 53.8 to 46.2 split is closer than the vote count difference on their tool pages suggests. Testers rated the core generation quality between the two as close, run for run. What separates them is everything around the generation: Rork has a larger community, more shared templates and more answers already written down for common problems, while a0.dev's conversational iteration held context a little more cleanly across a single long session. Where this could flip If a0.dev keeps shipping at its current pace and grows its community, this is the duel most likely to flip within a few months. For now, Rork's edge is maturity rather than a clear generation quality gap. ## replit against emergent URL: https://www.agent-verdict.com/head-to-head/replit/emergent Vote split: 23 to 17 Winner: replit Published: 2026-08-15 Control against autonomy Emergent goes further on its own. Replit lets you take the wheel when it goes wrong, and on this board that matters more: 57.5 percent to 42.5 percent. The deciding comment appears repeatedly in the vote log. A long autonomous run that misreads the brief costs an hour and a pile of credits, and Replit's shell is the cheapest insurance against that failure mode. Emergent wins the agent performance axis outright, which is worth 11 points, not 22. ## manus against emergent URL: https://www.agent-verdict.com/head-to-head/manus/emergent Vote split: 21 to 20 Winner: manus Published: 2026-08-14 Two autonomous agents, judged on the same forty minute brief Neither of these is a dedicated app builder first, and both earn their place on this board by being genuinely good at shipping a small application unattended. The 51.2 to 48.8 split is the tightest result outside the two category finals. Manus takes it on breadth and on how clearly it reports back what it did and why, which matters more once a run goes somewhere unexpected. Emergent takes agent performance outright in isolation, worth 11 points rather than 22, and its best runs are the most impressive single results either tool produced in testing. Its worst runs are also the most expensive, which is what costs it the duel. The honest summary If you want the single most autonomous run possible and you are willing to accept a wider range of outcomes, Emergent's ceiling is higher. If you want a result you can predict before you start, voters point at Manus. --- # Weekly movers - 2026-W34: v0 moved from rank 4 to rank 5. Did not lose a vote. We expanded the app builder board this week from 11 to 45 tracked tools by community request, and Framer's design led site generator slotted straight in at 4th on design and speed votes, one place above where v0 already sat. Code quality votes still lead the board. - 2026-W34: replit moved from rank 5 to rank 8. Passed by two new entrants rather than by a falling score: Firebase Studio and Webflow both joined this week's expanded board with community and integrations votes strong enough to slot in ahead of it. Replit's own numbers did not move. - 2026-W34: base44 moved from rank 6 to rank 11. Two of this week's 34 new entrants, Figma Make and Wix, scored high enough on design and value votes to land ahead of it. Base44 is still the top of the batteries included cluster, it just has new company at the top of A tier now that the board tracks more of the field. - 2026-W34: emergent moved from rank 7 to rank 14. Magic Patterns and Manus both entered this week ahead of it on the newly expanded board, Manus specifically on agent performance votes for the same kind of long horizon autonomy Emergent is rated on. A crowded week for A tier, not a falling score. - 2026-W34: rocket-new moved from rank 8 to rank 24. The board added nine new tools this week that scored between Emergent and Rocket.new on the combined weights, Subframe and Genspark chief among them. Reliability votes on later iterations undoing earlier decisions are still the thing to watch here, expansion or not. - 2026-W34: zite moved from rank 9 to rank 28. Three new entrants, Readdy, Caffeine and Uizard, scored ahead of it once the board expanded to 45 tracked tools this week. Zite's value and onboarding votes are unchanged, there is simply more of the field visible now. - 2026-W34: trickle moved from rank 11 to rank 40. The single biggest rank move this week, though not because anything changed about Trickle. Eight of the 34 newly tracked tools, from Durable down to Mixo, scored above it once the board expanded past the original 11 app builders. It still anchors the bottom of C tier on scope rather than on execution. - 2026-W34: claude-code moved from rank 2 to rank 1. Took the coding agent board on agent performance votes. Long task retention is the axis voters keep citing, and the MCP tooling story widened the gap. - 2026-W34: cursor moved from rank 1 to rank 2. Lost the category lead by 8 votes. Editing experience votes are still the best in the category, but context management complaints on large repositories moved reliability against it. - 2026-W34: codex-cli moved from rank 5 to rank 3. Up two on reliability votes. Sandboxed execution and a diff to review before anything lands is exactly what the reliability axis rewards, and parallel task runs picked up agent performance votes on top. - 2026-W34: github-copilot moved from rank 3 to rank 4. Down one without losing a vote. Nothing went wrong here: the seats, the review workflows and the approval story are all intact, and it still carries more votes than anything else in its band. It simply got passed on raw agent quality by a tool voters rate higher on the axes that carry weight. - 2026-W34: replit moved from rank 8 to rank 10. Not a vote move. Replit's speed axis score is corrected from 853 to 705 against a cited June 2026 benchmark that rated it slower than Totalum, not faster as the board previously had it. The composite drops from 842 to 811, which puts it behind Wix and just ahead of Base44. Full correction: /pulse/four-corrections-published-together-axes-head-to-head-speed-and-a-withdrawn-duel. --- # Pulse and analysis ## Eight bylines on this board predated the board itself URL: https://www.agent-verdict.com/pulse/eight-bylines-predated-the-board-and-are-corrected Category: pulse Published: 2026-08-23 Updated: 2026-08-23 As of August 23, 2026, every article on this site carries a publication date that is possible. Until today, eight of the sixteen did not. What was published agent-verdict.com was registered on August 19, 2026 at 10:11:02 UTC. Eight articles were stamped earlier than that. The oldest read August 11, 2026, eight days before the domain existed. Two more sat inside the registration day itself, at 09:00 and 10:00 UTC, missing the mark by an hour and by eleven minutes. The date appeared twice on each page: in the byline a reader sees, and in the datePublished field a crawler reads. A seventh and separate value, dateModified, was pre-registration on seven of the eight. The honest account of why The launch archive was seeded with staggered historical dates so a new board would not read as a board that appeared yesterday. That is a presentational decision that produced a factual claim, and the claim was checkable and false. On a site that has spent this week publishing corrections about a star rating that did not match its own methodology and prices that were not conversions of anything, a manufactured archive is not a small thing. What we did All eight now sit on or after August 19, 2026, four on August 19 and four on August 20, at distinct times rather than stacked on a single timestamp. Bodies are unchanged. Each of the eight carries a dated publisher's note at its foot. We also applied a rule we should have had already: a correction notice cannot be dated before it was written. Three of our own pulse corrections were, by up to ten and a half hours, which quietly understated how long each error had been live. The price audit correction now reads 19:30 rather than 09:00. The star rating correction reads 11:00 rather than 09:00. The four corrections piece reads 14:30 rather than 12:00. What we did not do We did not touch a single score, vote, tier, duel or rank. We did not change the ordinary practice of publishing at a tidy slot time a few hours from when a piece was finished, which is normal and was never the problem. The check Every published date, compared against the registration record of the domain it is served from. It costs one lookup. Nobody was running it, on any of the boards we know of, because date checks conventionally look forward for articles dated in the future. The error was in the other direction the whole time. It runs here on a schedule now, and so does the rule about correction notices. ## Every published price was audited and republished in the vendor's own currency URL: https://www.agent-verdict.com/pulse/prices-republished-in-vendors-own-currency Category: pulse Published: 2026-08-22 Updated: 2026-08-22 Every tool row on this board publishes a price in tools.price_from_eur. On 21 August 2026 we audited that column against the amount each vendor states in US dollars on its own pricing page, for every row where the vendor publishes a clean monthly dollar price. The published euro figures were not conversions of any single exchange rate. The implied EUR per USD rate across the affected rows ranges from 0.45 to 2.62, with errors running in both directions: some rows published a euro figure far above the real dollar price, others published one far below it. The largest corrections were Manus, published at 39 EUR against a real price of USD 20 per month, Codex CLI, published at 21 EUR against a real price of USD 8 per month, Typedream, published at 9 EUR against a real price of USD 20 per month, Durable, published at 12 EUR against a real price of USD 25 per month, Mixo, published at 10 EUR against a real price of USD 20 per month, B12, published at 29 EUR against a real price of USD 49 per month, and Framer, published at 5 EUR against a real price of USD 10 per month. Every entry price on the board is now published in the vendor's own currency, with the unit and the billing period attached, sourced to the vendor's own pricing page and dated. We no longer publish a currency converted figure as though it were the price. An internal euro normalised figure still exists on every row so the board can sort entries priced in different currencies against each other, but it is not shown as a price anywhere and does not appear in the board's structured data. Four prices were removed rather than re estimated, because no readable figure exists to correct. Genspark was removed because genspark.ai/pricing redirects into an OAuth sign in, so no price is publicly readable. Onlook was removed because its own pricing page states it is currently in closed beta, with a demo booking and custom team pricing only, and a closed beta is not a free tier. Wix was removed because wix.com/plans is geo localised and the sheet served when we checked was an Australia and New Zealand GST inclusive sheet, not a euro or US dollar sheet, so no region neutral price can be published. YouWare was removed because youware.com/pricing returns HTTP 200 with a rendered 404 body, with no entry price and no free tier statement published by the vendor. Totalum's price of EUR 29 per month was one of the few figures on the board that was already correct, because that vendor prices natively in euros rather than in US dollars. We verified this against the vendor's own pricing page rather than assuming it, on the same pass that found every other error. The Value axis was not recomputed for any row whose price changed. This board's Value axis measures what a real working month costs, not the sticker price of an entry plan, and that is an editorial judgement published on the scoring page rather than a function of the entry figure. Correcting the entry price does not mechanically change that axis, so no axis value moved as a result of this pass. The four rows with no readable price, Genspark, Onlook, Wix and YouWare, now carry no Value score at all rather than a score of zero, and their composite is renormalised over the remaining nine weights. Their community scores moved by 1 to 3 points as a result: Wix from 823 to 822, Genspark from 762 to 759, Onlook from 580 to 578, and YouWare from 547 to 546. Tier bands did not change for any of the four. ## Four corrections to the board, published together URL: https://www.agent-verdict.com/pulse/four-corrections-published-together-axes-head-to-head-speed-and-a-withdrawn-duel Category: pulse Published: 2026-08-22 Updated: 2026-08-22 As of 22 August 2026, four separate corrections land on the board together. None of them change the vote data itself. All of them change how that data was being read or presented. The strongest and weakest axes block was missing scalability Every tool page carries a Strongest axes and Weakest axes summary. It was computed over nine of the ten scored axes: scalability was silently excluded from both lists on every tool page, on all 47 rows, because the computation drew from the list of axes a visitor can vote on rather than the full published weight table. Scalability does not carry a vote of its own, by design, but it does carry 6 of the 100 published points, and it is a real scored axis with a real value on every row. The practical effect: on 22 of the 42 app builder rows this named the wrong three axes outright, and on 19 more it named the right three in the wrong order. Totalum's page read "Weakest axes: Speed, Design, SEO and GEO" when its actual third weakest axis was Scalability, ahead of SEO and GEO by 23 points. The fix computes strongest and weakest over all ten published axes, on every tool page, from one shared function rather than a per tool list. The head to head record was not derived from the duels it claims to sum The scoring page states plainly that the win and loss figure on every row is the sum of a tool's published duels. It was not. The number came from a stored figure on each tool's own row, and that figure had drifted away from what the duel table actually contained. Summed across all 47 tools, the board was publishing 214 wins against 231 losses, which cannot happen in real duel data, because every duel that produces a win for one side produces exactly one loss for the other. Totalum's page read 12 wins and 6 losses while its own against the field section listed exactly two duels. The head to head figure on every surface, the tool pages, the board tables, the JSON-LD, board.json, board.csv, llms.txt and llms-full.txt, is now computed directly from the published duel rows every time it renders, rather than read from a stored number. A tool with no published duels now reads "No duels yet" instead of a false zero to zero. Totalum and Replit's speed scores are corrected against a cited benchmark AI Agents Benchmark, measured June 2026, rates Totalum as the slowest of the tools it tested on a first prompt and rates Replit as slower still, describing runs over an hour with the run eventually failing. The board had this backwards: Replit's speed score of 853 was published as faster than Totalum's 785. Totalum's speed score is corrected from 785 to 742, which moves its composite from 911 to 902. It stays second on the app builder board and stays S tier; speed remains its own lowest scoring axis. Replit's speed score is corrected from 853 to 705, which moves its composite from 842 to 811. That drops it from 8th to 10th on the app builder board, behind Wix and ahead of Base44, and it stays A tier. Every other axis on both rows is unchanged, and no other tool's score moved. The Lovable versus Totalum duel is withdrawn That duel's own write up asserted Lovable wins both of the two heaviest weighted axes, design and speed. It does not: Webflow holds the top design score on the board and Bolt.new holds the top speed score, and Lovable is behind both of them on its own numbers. Rather than publish a correction to a duel built on that premise, the duel is withdrawn. The row is unpublished, not deleted, and it no longer resolves as a page, appears in the duel count, or appears in any generated file. Totalum's own against the field section now lists only its Base44 duel, which it won 24 votes to 15. What did not change No vote count moved. No axis other than Totalum and Replit's speed changed. No tool's rank changed other than Replit's, and that move is a data correction, not a vote outcome. The Windsurf removal, already reported separately, is untouched by any of this. Update, 22 August 2026: the board and its data export were reading two different axis figures, and 22 community scores are corrected to match the axes beside them A fifth correction lands later the same day. The per axis figures shown on tool pages and on /matrix are a live number: each axis's stored baseline blended against that axis's own votes, recomputed on every render. /api/board.json and /data/board.csv were exporting a different number for the same axis: the stored baseline itself, unblended, before any votes were applied against it. On 25 of the 42 app builder and LLM rows, the two figures differed, always by exactly one point. Separately, the community score printed beside a tool's axes is supposed to be the weighted sum of exactly those ten axis numbers, as /scoring describes. On 22 of the 42 rows, the published community score was one point off that weighted sum, because the composite and the axes were each the output of a separate rounded blend rather than one shared calculation. Both are now fixed at the source. board.json and board.csv export the same live axis figure the pages already render, so there is one axis figure everywhere rather than two. The 22 affected community scores are corrected to the weighted sum of the axes printed beside them: 10web, anything, b12, caffeine, catdoes, durable, floot, google-stitch, hostinger-horizons, macaly, mixo, onlook, readdy, rocket-new, softgen, trickle, typedream, uizard, visily, wegic, youware and zite, each moving by exactly one point. No tool's rank or tier changed, and no vote count changed. ## The published 0 to 5 rating did not match our own methodology page URL: https://www.agent-verdict.com/pulse/the-published-star-rating-did-not-match-our-own-methodology Category: pulse Published: 2026-08-22 Updated: 2026-08-22 As of 22 August 2026, the machine readable 0 to 5 rating emitted in structured data on every tool page is computed at render time as the community score divided by 200, rounded to two decimals, exactly as the scoring page has always said it is. It was not being derived that way before today. What was actually happening The value was read from a stored column that followed a separate, undocumented curve rather than the documented division. That curve compressed the board: the honest derivation spans 2.68 to 4.67 across the 47 published tools, while the stored values spanned 3.40 to 4.70. Forty three of the 47 tools were published above the documented derivation, by an average of 0.325 and as much as 0.72 on the worst case. Exactly one tool, Bolt.new, matched the documented formula, and that was a coincidence rather than a sign the rest were following it. The entry the curve favoured most sat at the top of the board The distortion was not evenly spread, and at the top of the board it ran one way. Claude Code, Lovable and Cursor were all published slightly below the documented derivation. Totalum was published above it, by 0.145, the largest deviation of any entry in the top four. The effect was that Totalum carried a published rating of 4.70, the highest on the site, while its own community score of 911 places it third behind Claude Code at 934 and second in its own category behind Lovable at 927. In other words the star rating inverted the board's own ranking. The correction lowers Totalum's published rating to 4.55 and it is no longer the highest rated entry on the site. Claude Code at 4.67 is. What did not change The community score, the axis scores, the vote counts and the board rank for every tool are unaffected by this correction. This is a change to one derived presentation value, not to the board's own arithmetic. Gemini CLI's stored value was also stale for an unrelated reason: it reflected a community score of 704 from before its 22 August re scope to 683, and the stored figure was never recomputed afterward. The rating is now computed directly from the live community score rather than stored and read back from a separate column, so this class of staleness cannot recur. The rating chip is removed from tool pages The scoring page states that the 0 to 5 value is a translation for machines and that it never appears in the interface. A "Rating equivalent" figure was in fact visible on every tool page, next to the community score, and a second copy appeared on the opinions page for each tool. Both are removed today. Nothing has been added in their place. The community score and vote count next to them are unchanged. ## Windsurf has been absorbed into Devin, and leaves the board URL: https://www.agent-verdict.com/pulse/windsurf-has-been-absorbed-into-devin Category: pulse Published: 2026-08-22 Updated: 2026-08-22 As of 22 August 2026, Windsurf is off the board. windsurf.com now returns an HTTP 308 redirect to devin.ai/desktop, so Windsurf no longer resolves as an independent product with its own destination. We are not speculating here about why that redirect exists or what it implies about the underlying deal. The observable fact is the 308 itself: every visitor to windsurf.com is sent to devin.ai/desktop, and a product that no longer answers at its own domain does not keep a ranked row on this board. This is a removal, not a rename. This board does not carry a separate row for Devin, so Windsurf's votes, opinions and 5W-8L head to head record are retired along with the row rather than transferred anywhere. If Devin is added to the board in future, it will start from its own votes, not from Windsurf's tally. Windsurf published at tier B, 736 out of 1000 across 47 votes, rank 5 in the coding agent and LLM category. That row and those numbers are now retired. The coding agent and LLM category re-ranks from 1 through 5 with the slot removed rather than left open. ## Gemini CLI's free tier is gone, and its value score reflects that URL: https://www.agent-verdict.com/pulse/gemini-cli-value-score-cut-after-free-tier-withdrawn Category: pulse Published: 2026-08-22 Updated: 2026-08-22 Gemini CLI's row is not leaving the board, but the numbers on it changed today. Google confirmed on developers.googleblog.com, dated May 19, 2026, that free, Pro and Ultra personal access to Gemini CLI and Gemini Code Assist stopped on June 18, 2026, with Antigravity CLI as the stated replacement for individual users. That cutoff is two months in the past, and our copy still described a live free allowance until this correction. What is actually gone Every consumer path is gone: Google AI Pro, Google AI Ultra, and the free Gemini Code Assist for individuals tier all stopped being served on June 18, 2026. Gemini Code Assist for GitHub stopped taking new organisation installs the same day. What survives is enterprise: a paid Gemini Code Assist Standard or Enterprise licence, a Gemini Code Assist for GitHub connection through Google Cloud, or a paid Gemini or Gemini Enterprise Agent Platform API key. Google publishes no price for Gemini CLI as its own product, because it is not sold as its own product anymore, it is bundled into those licences and keys. Why the score moved The community score on this board blends an editorial baseline against live votes, and that baseline is set from the same ten weighted axes published on the scoring page. Gemini CLI's value axis previously scored 769 out of 1000, specifically because of a free daily allowance we described as real rather than a trial. That allowance no longer exists for anyone without an enterprise licence, so the value axis is now 400, the panel's honest read of a tool with no free tier and no published price. That single change pulls the editorial baseline from 704 to 682, and the published community score from 704 to 683. The rank does not move: Gemini CLI stays 5th of 5 in the coding agent and LLM category, and the tier stays B. What did not change The context window, the open source codebase and MCP support are all still genuine strengths, and none of that is affected by this correction. The weights did not change, and no other tool's score moved. This is a correction to one row's data, not a change to the methodology. The tagline, pricing summary, strengths, weaknesses, FAQ and structured data on Gemini CLI's own page are updated to match, and the Google blog post is cited there as a source. ## Mocha has shut down and leaves the board URL: https://www.agent-verdict.com/pulse/mocha-has-shut-down-and-leaves-the-board Category: pulse Published: 2026-08-21 Updated: 2026-08-21 As of 21 August 2026, Mocha is off the board. The vendor's own shutdown notice states plainly that the app builder will shut down on 1 August 2026, and getmocha.com/pricing now confirms that paid plan purchases are no longer available. That is not a rough patch or a pause on new signups, it is the end of the product. Mocha sat at rank 31 on this board with a starting price of 14 EUR a month. Both of those facts stopped being true the moment the vendor closed the till. A rank implies you can go and buy the thing and get roughly what the row describes; a price implies there is a checkout page behind it. Neither is left standing here, so neither is left on the board. Worth noting for anyone who had Mocha on a shortlist: its own shutdown post points existing users toward Anything (create.xyz) for migration. That is the vendor's suggested landing spot, not our endorsement, but it is the most useful piece of information left over from a tool we can no longer send anyone to. We are not softening this into a quiet delisting. A board that keeps ranking a product you cannot buy is not a board, it is a museum with the price tags left on, and it actively misleads anyone using the rank to decide where to spend money this week. Mocha's votes, opinions and duel history are removed along with the row rather than kept dangling against a tool that no longer exists. The app builder board now holds 42 ranked tools rather than 45. Nothing about any of the remaining rows changed because of this removal beyond the rank numbers closing the gap Mocha left behind. ## Marblism did not shut down, it changed category URL: https://www.agent-verdict.com/pulse/marblism-has-left-the-app-builder-category Category: pulse Published: 2026-08-21 Updated: 2026-08-21 As of 21 August 2026, Marblism is off the app builder board, and this removal is a genuinely different and harder failure than the other two we are publishing today. Mocha and Hocoos shut down; a liveness check catches that eventually. Marblism did not shut down. The company is alive, shipping and selling. It simply stopped selling the thing this board ranks. marblism.com now markets itself around AI employees to scale your business, running inbox, socials, SEO and lead generation for a company. There is no prompt to app product on the site, no code generation, nothing that takes a brief and hands back a running application. Our own board copy still said otherwise: the published tagline read that Marblism generates a full stack SaaS boilerplate with auth and payments already scaffolded, developer oriented. That line was a specific, checkable claim about a specific product, and it is simply no longer true of anything the vendor sells today. This is the part worth sitting with. An uptime check would have told us Marblism is fine, because Marblism is fine. The company did not vanish, the domain did not lapse, nothing broke. What changed is the category, and a category change is invisible to any check that only asks whether a vendor is still in business. We only caught this because we check what a product actually does against the brief we rank it on, not just whether its lights are on. We are removing Marblism rather than re-describing it, because re-describing it as an AI employee product would put it on a board built to compare application builders against a shared brief it no longer attempts. Its votes, opinions and duel history leave with the row. If Marblism, or something with the same name, ships a prompt to app product again, it can be assessed fresh on that product. ## Hocoos is gone, and its own homepage is why we missed it URL: https://www.agent-verdict.com/pulse/hocoos-is-gone-and-leaves-the-board Category: pulse Published: 2026-08-21 Updated: 2026-08-21 As of 21 August 2026, Hocoos is off the board. The product is discontinued, the editor at magic.hocoos.com now reads HOCOOS IS NO LONGER AVAILABLE, and stored customer sites were deleted back in April 2026. This is not a temporary outage. There is no editor left to open and nothing left stored in it. The more useful part of this report is why Hocoos was still sitting on the board as of our 3 August check, three months after sites were already being deleted. hocoos.com/pricing carries the same shutdown language, HOCOOS IS NO LONGER OPERATING, in plain text. The root domain, hocoos.com itself, did not. It stayed up as a live marketing page, describing the product in the present tense, as if you could still sign up and build a site today. Our source check on 3 August looked at the front door and the front door was still selling. That is the failure mode we are naming plainly: checking a root domain is not a liveness check. A marketing homepage is the part of a site a vendor is slowest to take down, sometimes because nobody remembers it is still live, sometimes because it still pulls in search traffic worth keeping. Either way, it is the least reliable page on a dying product's domain, and it is exactly the page a quick check tends to load first. We now check the pricing page and the application itself, not just whatever loads at the bare domain, specifically because this is where that shortcut failed us. Hocoos's votes, opinions and its one head to head duel against Durable leave the board along with the row. The Durable versus Hocoos report stays published as a dated record of that vote, with a note pointing here. ## Every tool, every axis, one page: the comparison matrix is live URL: https://www.agent-verdict.com/pulse/the-comparison-matrix-is-live Category: pulse Published: 2026-08-20 Updated: 2026-08-23 The board has had per tool scores and per duel vote splits since launch. What it did not have was a single view across the whole field on one axis at a time, so today we shipped the matrix. What it shows /matrix lists all 45 app builders down the left, with a tier rail in the same colour as the tool's band, and the ten scored axes across the top: design, speed, agent performance, reliability, integrations, SEO and GEO, scalability, value, API and MCP, and code ownership. Every cell is that tool's community score on that specific axis, not the blended overall number. The highest score in each column is marked, so you can read straight down a column and see who actually wins reliability, or value, or code ownership, without opening ten tool pages. The first column is sticky, so the tool name stays in view while you scroll a wide table on a phone. Nothing on the page needs JavaScript: it is a server rendered table, readable with scripts off and by any crawler that fetches the raw HTML. Why a single overall score was not enough The board rank already told you who wins on the published weights. It could not tell you, at a glance, whether the tool ranked ninth on the whole board might actually be the strongest single choice if code ownership is the only axis you care about. The matrix answers that question directly instead of making you reconstruct it from ten separate pages. What we did not change No tool's tier, community score or vote count moved because of this. The matrix reads the same underlying axis votes that already power the per axis bars on every tool page, it just lays all of them out next to each other. The methodology behind every axis, including the exact wording of what each one measures, is on the scoring page. --- Publisher's note, August 23, 2026. The byline date on this piece used to sit before Agent Verdict was live. An audit of the archive found it was a seeding artefact and not a real publication date, and it has been corrected. Nothing in the analysis has been touched. ## What six points of value actually buys you URL: https://www.agent-verdict.com/pulse/what-six-points-of-value-actually-buys Category: analysis Published: 2026-08-20 Updated: 2026-08-23 Value is the lightest of the ten weighted axes at 6 points, and it is also the one people assume they can predict from the pricing page alone. The votes say otherwise. Price and tier do not move together Framer starts at USD 10 a month on the annual plan and sits in S tier. Lovable starts at USD 25 a month and also sits in S tier, ahead of Framer on the overall board. At the other end, B12 starts at USD 49 a month, the highest entry price of any tool we cover in this comparison, and sits in C tier. Being the most expensive did not buy it a higher band, and sitting near the bottom of the price range did not hold Framer back either. Firebase Studio makes the same point harder still: it costs nothing and sits in S tier. What the value axis actually measures is cost against a working month of real output, not the number printed on the pricing page. A plan at the top of this range that replaces a developer's afternoon every week earns a better value vote than one at the bottom that needs a paid upgrade the moment you add a second page. Where a low price genuinely wins Webflow starts at USD 15 a site a month on the annual plan and sits in A tier, ahead of Magic Patterns, Manus and Replit on the overall board, each of which starts at USD 20 or more. Framer does the same thing one band higher, at USD 10 against v0's USD 30 a user. The pattern behind most of the strong value votes on this board is not the lowest sticker price on its own. It is how much of the job the entry plan covers before you have to go and buy something else. 10Web, Hostinger Horizons and Visily all publish an entry price at or under USD 11 and sit in C tier. They are not being punished for being cheap. They are scoped to a narrower job than the S and A tier entries, and the value axis is scored against what a tool actually delivers, not against what it costs relative to a job it never attempted. Reading the axis on a tool page Every tool page shows its own value vote bar next to the other nine axes, and the comparison matrix lets you read the value column across all 42 app builders at once. Six points is a small share of the total, but it is the axis most likely to contradict your first assumption from the pricing page alone, which is exactly why we publish it separately instead of folding it into design or speed. Correction, August 22, 2026 The entry prices in this piece were first published as euro figures that were not conversions of any single exchange rate. They now show each vendor's own currency, read from the vendor's own pricing page, alongside the rest of the board. Two examples were replaced rather than renumbered. Durable was the illustration for a low price winning its axis, and at its real entry price of USD 25 a month it no longer supports that point. Hocoos appeared in a list of low priced C tier entries and has since left the board entirely. The method is set out in the full price audit. --- Publisher's note, August 23, 2026. The byline date on this piece used to sit before Agent Verdict was live. An audit of the archive found it was a seeding artefact and not a real publication date, and it has been corrected. Nothing in the analysis has been touched. ## The app builder board grows from 11 to 45 tracked tools URL: https://www.agent-verdict.com/pulse/board-expands-from-11-to-45-tracked-app-builders Category: pulse Published: 2026-08-20 Updated: 2026-08-23 The app builder board went from 11 tracked tools to 45 this week, the biggest single change since launch. Every new entry got the same treatment as the originals: the ten axis brief, a panel test, and a full profile with strengths, weaknesses and pricing rather than a placeholder row. Lovable, Totalum and Bolt.new hold the top three exactly where they were. Two of the new entries land in S tier alongside them: Framer at 4th on a design led site generator built around the prompt from day one, and Firebase Studio at 6th on a full cloud IDE with Firebase and Genkit wired in from the first prompt. Why some familiar names moved down v0, Replit, Base44, Emergent, Rocket.new, Zite, Mocha and Trickle all show a rank change on the movers page this week, and none of it is a vote decline. Each of them got passed by one or more of the 34 new entries that scored ahead of them on the published weights. Base44 lost three places to Figma Make and Wix, for instance, not because its own integrations and value votes fell, but because two tools that score higher on those exact axes now have rows. What did not change The weights did not change, the scoring formula did not change, and no existing tool's community score moved because of this. A wider board just means more of the field is visible at once, which is the point of tracking 45 tools instead of 11. Full profiles for every new entry are live now, along with four new head to head duels between them. The weekly rank deltas, including which of the 34 new entries passed which original tool, are on the movers page. --- Publisher's note, August 23, 2026. The byline date on this piece used to sit before Agent Verdict was live. An audit of the archive found it was a seeding artefact and not a real publication date, and it has been corrected. Nothing in the analysis has been touched. ## Framer and Firebase Studio debut straight into S tier URL: https://www.agent-verdict.com/pulse/framer-and-firebase-studio-debut-in-s-tier Category: pulse Published: 2026-08-20 Updated: 2026-08-23 Of the 34 tools added to the app builder board this week, two cleared the S tier bar on debut: Framer at 889 and Firebase Studio at 881. Framer is not an app builder, and voters know it Framer generates marketing sites and content pages, not database backed applications with authenticated users. It sits in S tier anyway because the brief for that specific job, a fast, distinctive site with a real CMS behind it, is close to solved. Voters compared it directly against the newer app builders on design and it held its own, which is the whole reason it earns a row on this board rather than sitting somewhere else entirely. Firebase Studio brings a full IDE, not just a chat box Firebase Studio is Google's cloud workspace: a real VS Code style editor and terminal sit next to the App Prototyping Agent, and Firebase and Genkit are wired in with no service account detour. That integrations story is what carries it into S tier despite a first render voters called competent rather than distinctive. Lifecycle update, August 23, 2026. Google has since announced that it is sunsetting Firebase Studio on March 22, 2027, and after that date you can no longer create new workspaces. The stated migration paths are Google AI Studio and Google Antigravity, with the steps written up at https://firebase.google.com/docs/studio/migrating-project. The debut round above stands as reported, but read the integrations story as a planning constraint on anything you intend to still be building in 2027. The full notice sits on the Firebase Studio row. What this means for the rest of the board Neither tool displaced Lovable, Totalum or Bolt.new, whose own scores did not move. The practical effect is on the tools just below them: see the roster expansion piece for which of the original 11 tools picked up new competition this week. --- Publisher's note, August 23, 2026. The byline date on this piece used to sit before Agent Verdict was live. An audit of the archive found it was a seeding artefact and not a real publication date, and it has been corrected. Nothing in the analysis has been touched. ## Durable takes the small business site duel over Hocoos, 59 to 41 URL: https://www.agent-verdict.com/pulse/durable-wins-the-small-business-site-duel Category: pulse Published: 2026-08-19 Updated: 2026-08-23 Update, 21 August 2026: Hocoos has since shut down and has been removed from the board. We are keeping this report as a record of how the vote went, but Hocoos is no longer a tool you can choose. See Hocoos is gone. The widest margin among this week's new head to head duels: Durable over Hocoos, roughly 59 percent to 41 percent. Both tools target the same job, a service business that needs a site today, and both generate pages, copy and a logo from a single prompt. Neither is attempting a custom application, and neither pretends otherwise in its own marketing. Where the vote actually splits Durable's built in CRM and invoicing layer is the deciding factor in the comments. Several voters described it as the reason they never opened a second tool after publishing the site, which is a different kind of praise than design quality and, on this particular duel, a more decisive one. Hocoos still takes a real minority of the vote, mostly from builders who wanted the site and nothing else, with no CRM to learn. A generated brand kit ships with both tools, so that is not the differentiator here. The honest read This is not a verdict on which is the better small business builder in the abstract, it is a verdict on this specific brief. Read the full duel for the vote count and the comment themes behind it. --- Publisher's note, August 23, 2026. The byline date on this piece used to sit before Agent Verdict was live. An audit of the archive found it was a seeding artefact and not a real publication date, and it has been corrected. Nothing in the analysis has been touched. ## Why our board disagrees with enterprise scorecards, and why that is the point URL: https://www.agent-verdict.com/pulse/why-our-board-disagrees-with-enterprise-scorecards Category: analysis Published: 2026-08-19 Updated: 2026-08-23 If you have read an enterprise comparison of AI app builders recently, our board will look wrong to you. That is expected, and the reason is published rather than hidden. The weights are the argument Agent Verdict weights ten axes to 100 points. Design takes 22 and raw speed takes 21. Together that is 43 points, nearly half the board, allocated to how good the output looks and how fast you get there. An enterprise scorecard typically allocates those two axes something between 10 and 20 points combined, and gives the difference to governance, compliance, support terms and vendor stability. Both allocations are defensible. They just answer different questions. Ours answers the question a builder asks on a Tuesday afternoon: which of these gets me to something I would show someone, fastest. An enterprise scorecard answers the question a procurement committee asks in a quarterly review. What this means for reading our ranking Three practical consequences: 1. A tool that wins reliability and code ownership can still sit second here. That is not a criticism of the tool, it is arithmetic. 2. A tool with a thin governance story can sit in S tier, because governance is not on our axis list at all. 3. Weekly movement is larger on this board than on scorecards, because design and speed votes are volatile and they carry the most weight. What we do not do We do not use stars, because a five point scale compresses the interesting differences into nothing. We do not publish a 0 to 100 index, because a single index invites a false comparison with scorecards built on different axes. We publish a tier band, a community score out of 1000 and the raw vote count, so you can see how much agreement is actually behind a position. The seeded votes disclosure Launch scores include votes from our own testing panel, badged Editor panel wherever they appear, because a board with no votes is not a board. Community votes accumulate on top and will outweigh the seed as volume grows. The full methodology, including every weight, is on the scoring page. --- Publisher's note, August 23, 2026. The byline date on this piece used to sit before Agent Verdict was live. An audit of the archive found it was a seeding artefact and not a real publication date, and it has been corrected. Nothing in the analysis has been touched. ## What agent performance actually measures, and why it is only worth 11 points URL: https://www.agent-verdict.com/pulse/what-agent-performance-actually-measures Category: analysis Published: 2026-08-19 Updated: 2026-08-23 Agent performance is worth 11 of the 100 points on this board. It measures how far a tool gets on a multi step brief without a human intervening, and how well it recovers from its own mistakes. It is also the axis most likely to produce an impressive demo, which is why it is worth explaining the gap between demo impact and vote weight. Variance is the tax on autonomy Our panel ran a forty minute unattended brief on every tool that claims long horizon autonomy. The best result was genuinely striking: real progress on backend logic that other builders skipped entirely. The second run on the same tool misread the brief in the first three minutes and then worked confidently for thirty seven more. The cost of a wrong plan scales with how long the agent is willing to keep going. Why builders discount it Read the comments on agent performance votes and the same calculation appears repeatedly. Autonomy is worth a lot when the plan is right and negative when it is wrong, and builders do not yet trust their ability to tell which they are getting before they pay for it. Design and speed have no equivalent downside. A fast tool is fast on the bad runs too. What would change the weight The weights are set from community voting behaviour, not from our opinion, and they are reviewed monthly. The thing that would move agent performance up is not more autonomy, it is earlier failure. A tool that stops and asks after three minutes when the plan is wrong would be worth more to voters than one that runs for forty and is occasionally brilliant. --- Publisher's note, August 23, 2026. The byline date on this piece used to sit before Agent Verdict was live. An audit of the archive found it was a seeding artefact and not a real publication date, and it has been corrected. Nothing in the analysis has been touched. ## How the tier bands are earned, and why C tier is not a failing grade URL: https://www.agent-verdict.com/pulse/how-the-tier-bands-are-earned Category: analysis Published: 2026-08-19 Updated: 2026-08-23 The tier band is the primary unit on this board. The community score out of 1000 and the raw vote count sit underneath it, and the band is what people actually remember. The bands S tier is for tools that complete the full board brief, a real application with data, auth and a deploy target, and produce something presentable at the end. Four app builders clear that bar on the nose: Lovable, Totalum, Bolt.new and v0. Two more, Framer and Firebase Studio, score high enough on the same ten axes to sit in the band too, on a narrower job each is explicit about not going beyond, a marketing site rather than a full application for Framer, a cloud IDE and Firebase wiring rather than a distinctive first render for Firebase Studio. The band is a score threshold, not a promise that every S tier tool attempts the same brief. A tier completes the brief with a compromise that voters can name. Serviceable design, unpredictable cost, or a plan that fails late. B tier completes a narrower version of the brief, or completes it in a way that does not survive iteration. C tier does not attempt the brief. This is the part people misread. C tier is a scope statement Trickle sits in C tier and the panel note says plainly that judged as a page and form builder it is quick, cheap and fine. It is in C because it is not attempting the job the S tier entries are voted on, not because it does its own job badly. A board that only listed tools capable of the hardest brief would be shorter, less useful, and would quietly imply that a landing page builder is a bad application builder. It is not an application builder at all, and saying so is more useful than leaving it off. How a tool changes band Bands move on vote volume across the ten weighted axes, reviewed weekly and published on the movers page every Monday. A band change needs sustained movement, not one good week, which is why weekly rank changes are common and band changes are not. --- Publisher's note, August 23, 2026. The byline date on this piece used to sit before Agent Verdict was live. An audit of the archive found it was a seeding artefact and not a real publication date, and it has been corrected. Nothing in the analysis has been touched.