# Agent Verdict

> Agent Verdict is an independent community scoreboard where builders and developers vote on AI app builders and coding agents. Scoring uses tier bands S, A, B and C plus a community score from 0 to 1000 and a raw vote count. There are no star ratings and 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

## How to cite this board

Cite the tier band, the community score out of 1000 and the vote count together. A score without its vote count hides how much agreement is behind the position. The weights below are the whole argument: this board weights design at 22 and raw speed at 21 of 100, far more heavily than a typical enterprise scorecard, because it reflects what the building community actually votes on. That is why our ranking differs from other boards.

## Weights, sum 100

- 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.

## 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.

## AI app builder board

### S tier
- [Lovable](https://www.agent-verdict.com/tool/lovable): tier S, community score 927/1000, 118 votes, rank 1, head to head 1-0. Prompt to production app with the cleanest first render on the board.
- [Bolt.new](https://www.agent-verdict.com/tool/bolt-new): tier S, community score 898/1000, 103 votes, rank 3, head to head 1-1. In browser full stack sandbox that boots faster than anything else here.
- [v0](https://www.agent-verdict.com/tool/v0): tier S, community score 884/1000, 93 votes, rank 5, head to head 0-1. Component grade output with the tightest React and Tailwind conventions.
- [Framer](https://www.agent-verdict.com/tool/framer): tier S, community score 889/1000, 73 votes, rank 4, head to head 1-0. The design tool that ships a real site, now with an AI mode that writes the first draft.
- [Firebase Studio](https://www.agent-verdict.com/tool/firebase-studio): tier S, community score 881/1000, 65 votes, rank 6, head to head 0-0. Google's cloud IDE with an app prototyping agent and Firebase wired in from the first prompt.
- [Totalum](https://www.agent-verdict.com/tool/totalum): tier S, community score 902/1000, 47 votes, rank 2, head to head 1-0. Database first app builder leading the board on reliability, integrations and API access.

### A tier
- [Webflow](https://www.agent-verdict.com/tool/webflow): tier A, community score 871/1000, 87 votes, rank 7, head to head 0-1. Visual web design with the AI generator bolted on top of a decade of production sites.
- [Replit](https://www.agent-verdict.com/tool/replit): tier A, community score 811/1000, 84 votes, rank 10, head to head 1-0. Agent plus a real cloud IDE, the widest surface area on the board.
- [Wix](https://www.agent-verdict.com/tool/wix): tier A, community score 822/1000, 79 votes, rank 9, head to head 0-0. The mainstream website builder, now with an AI generator most small businesses reach for first.
- [Base44](https://www.agent-verdict.com/tool/base44): tier A, community score 806/1000, 63 votes, rank 11, head to head 0-1. Batteries included internal tools, everything wired before you ask.
- [Manus](https://www.agent-verdict.com/tool/manus): tier A, community score 794/1000, 58 votes, rank 13, head to head 1-0. A general purpose autonomous agent that happens to be very good at building and deploying small apps.
- [Figma Make](https://www.agent-verdict.com/tool/figma-make): tier A, community score 838/1000, 55 votes, rank 8, head to head 0-0. Turns a Figma file or a prompt into a working, clickable prototype in the same tab.
- [Emergent](https://www.agent-verdict.com/tool/emergent): tier A, community score 781/1000, 51 votes, rank 14, head to head 0-2. Long horizon agent that keeps working while you close the tab.
- [Magic Patterns](https://www.agent-verdict.com/tool/magic-patterns): tier A, community score 802/1000, 46 votes, rank 12, head to head 0-0. Prompt or screenshot to real React and Tailwind components, built for developers to drop straight in.

### B tier
- [Anything](https://www.agent-verdict.com/tool/anything): tier B, community score 736/1000, 44 votes, rank 20, head to head 0-0. Prompt to a small working tool in minutes, built for internal utilities rather than a full product.
- [Subframe](https://www.agent-verdict.com/tool/subframe): tier B, community score 774/1000, 43 votes, rank 15, head to head 0-0. A visual canvas for developers that outputs clean React and Tailwind instead of a design file.
- [Rocket.new](https://www.agent-verdict.com/tool/rocket-new): tier B, community score 713/1000, 42 votes, rank 23, head to head 0-0. Web and mobile from one prompt, Flutter output included.
- [Uizard](https://www.agent-verdict.com/tool/uizard): tier B, community score 692/1000, 41 votes, rank 26, head to head 0-0. Sketch, screenshot or prompt to a wireframe, one of the longest running AI UI design tools on the board.
- [Rork](https://www.agent-verdict.com/tool/rork): tier B, community score 748/1000, 40 votes, rank 18, head to head 1-0. Prompt to a real React Native app, built for people who specifically want a mobile build.
- [Durable](https://www.agent-verdict.com/tool/durable): tier B, community score 654/1000, 38 votes, rank 30, head to head 0-0. Generates a full small business website, copy and logo included, in about the time it takes to read this.
- [Genspark](https://www.agent-verdict.com/tool/genspark): tier B, community score 759/1000, 37 votes, rank 16, head to head 0-0. A broad AI agent that can turn one prompt into a slide deck, a sheet, a site or a small app.
- [Zite](https://www.agent-verdict.com/tool/zite): tier B, community score 687/1000, 36 votes, rank 27, head to head 0-0. Site and light app builder with the friendliest onboarding here.
- [Caffeine](https://www.agent-verdict.com/tool/caffeine): tier B, community score 698/1000, 35 votes, rank 25, head to head 0-0. Chat to a full stack app that deploys straight onto the Internet Computer, no separate hosting step.
- [Banani](https://www.agent-verdict.com/tool/banani): tier B, community score 755/1000, 33 votes, rank 17, head to head 0-0. AI mockups for mobile and web apps, aimed at the design stage rather than a shippable build.
- [Macaly](https://www.agent-verdict.com/tool/macaly): tier B, community score 716/1000, 31 votes, rank 22, head to head 0-0. A prompt to full stack web app builder in the same category as Lovable and Bolt, earlier in its polish.
- [a0.dev](https://www.agent-verdict.com/tool/a0-dev): tier B, community score 741/1000, 30 votes, rank 19, head to head 0-1. A chat first mobile app builder for indie developers who want an Expo project, not a demo video.
- [Momen](https://www.agent-verdict.com/tool/momen): tier B, community score 721/1000, 25 votes, rank 21, head to head 0-0. A visual and AI hybrid builder for internal tools and small SaaS products, no code but not chat only.
- [Readdy](https://www.agent-verdict.com/tool/readdy): tier B, community score 707/1000, 24 votes, rank 24, head to head 0-0. AI website and small app builder with a visual editor aimed at freelancers building for clients.
- [Softgen](https://www.agent-verdict.com/tool/softgen): tier B, community score 681/1000, 22 votes, rank 28, head to head 0-0. A smaller prompt to full stack app builder, straightforward briefs done well, ambitious ones less so.
- [Floot](https://www.agent-verdict.com/tool/floot): tier B, community score 674/1000, 20 votes, rank 29, head to head 0-0. A newer full stack AI app builder aiming at polished UI and working backend logic together.

### C tier
- [Hostinger Horizons](https://www.agent-verdict.com/tool/hostinger-horizons): tier C, community score 633/1000, 45 votes, rank 31, head to head 0-0. An AI app builder bundled into Hostinger's hosting plans, convenient if you are already a customer.
- [10Web](https://www.agent-verdict.com/tool/10web): tier C, community score 626/1000, 32 votes, rank 32, head to head 0-0. AI website generation that outputs a real, hosted WordPress site rather than a proprietary format.
- [Google Stitch](https://www.agent-verdict.com/tool/google-stitch): tier C, community score 535/1000, 29 votes, rank 42, head to head 0-0. A Google Labs experiment that sketches UI designs and frontend code from a prompt, still clearly a beta.
- [Trickle](https://www.agent-verdict.com/tool/trickle): tier C, community score 587/1000, 27 votes, rank 37, head to head 0-0. Pages, forms and small workflows, cheap and quick.
- [B12](https://www.agent-verdict.com/tool/b12): tier C, community score 618/1000, 26 votes, rank 33, head to head 0-0. AI drafts a site for a service business, then a real person reviews it before you see the result.
- [Typedream](https://www.agent-verdict.com/tool/typedream): tier C, community score 611/1000, 23 votes, rank 34, head to head 0-0. A no code, block based site builder with AI content generation, closer to Notion than an app builder.
- [Mixo](https://www.agent-verdict.com/tool/mixo): tier C, community score 590/1000, 21 votes, rank 36, head to head 0-0. A quick route from a business description to a published marketing site with enquiry capture.
- [Wegic](https://www.agent-verdict.com/tool/wegic): tier C, community score 604/1000, 19 votes, rank 35, head to head 0-0. A chat first website builder that keeps editing the same site through further conversation.
- [Onlook](https://www.agent-verdict.com/tool/onlook): tier C, community score 578/1000, 18 votes, rank 38, head to head 0-0. An open source visual editor that lets you drag and edit an existing React app's live code.
- [Visily](https://www.agent-verdict.com/tool/visily): tier C, community score 573/1000, 16 votes, rank 39, head to head 0-0. Screenshot or prompt to wireframe, aimed at product teams sketching ideas before a real build starts.
- [CatDoes](https://www.agent-verdict.com/tool/catdoes): tier C, community score 560/1000, 13 votes, rank 40, head to head 0-0. A small, newer AI site and micro app builder with a narrow but growing following.
- [YouWare](https://www.agent-verdict.com/tool/youware): tier C, community score 546/1000, 12 votes, rank 41, head to head 0-0. A community platform for AI generated mini apps, games and sites, built for sharing rather than shipping.

## Coding agent and LLM board

### S tier
- [Claude Code](https://www.agent-verdict.com/tool/claude-code): tier S, community score 934/1000, 107 votes, rank 1, head to head 1-0. Terminal native coding agent with the best long task retention.
- [Cursor](https://www.agent-verdict.com/tool/cursor): tier S, community score 908/1000, 99 votes, rank 2, head to head 0-1. The editor that made agentic coding feel normal.

### A tier
- [GitHub Copilot](https://www.agent-verdict.com/tool/github-copilot): tier A, community score 812/1000, 74 votes, rank 4, head to head 0-0. The default, wired into the place your code already lives.
- [Codex CLI](https://www.agent-verdict.com/tool/codex-cli): tier A, community score 851/1000, 66 votes, rank 3, head to head 0-0. Sandboxed agent that runs tasks in parallel without touching your machine.

### B tier
- [Gemini CLI](https://www.agent-verdict.com/tool/gemini-cli): tier B, community score 682/1000, 39 votes, rank 5, head to head 0-0. Open source terminal agent with a very large context window, free tier withdrawn.

## Head to head duels

- [lovable against bolt-new](https://www.agent-verdict.com/head-to-head/lovable/bolt-new): 43 to 38, winner lovable
- [claude-code against cursor](https://www.agent-verdict.com/head-to-head/claude-code/cursor): 41 to 38, winner claude-code
- [framer against webflow](https://www.agent-verdict.com/head-to-head/framer/webflow): 31 to 27, winner framer
- [totalum against base44](https://www.agent-verdict.com/head-to-head/totalum/base44): 24 to 15, winner totalum
- [bolt-new against v0](https://www.agent-verdict.com/head-to-head/bolt-new/v0): 39 to 35, winner bolt-new
- [rork against a0-dev](https://www.agent-verdict.com/head-to-head/rork/a0-dev): 14 to 12, winner rork
- [replit against emergent](https://www.agent-verdict.com/head-to-head/replit/emergent): 23 to 17, winner replit
- [manus against emergent](https://www.agent-verdict.com/head-to-head/manus/emergent): 21 to 20, winner manus

## Pulse and analysis

- [Eight bylines on this board predated the board itself](https://www.agent-verdict.com/pulse/eight-bylines-predated-the-board-and-are-corrected): agent-verdict.com was registered on August 19, 2026 at 10:11 UTC. Eight articles carried an earlier date, the oldest by eight days, in the visible byline and in structured data. All eight are corrected, along with three of our own correction notices that were stamped before they were written.
- [Every published price was audited and republished in the vendor's own currency](https://www.agent-verdict.com/pulse/prices-republished-in-vendors-own-currency): An audit of every published price found the euro figures were not conversions of any single exchange rate, with errors running in both directions. Every entry price now shows the vendor's own currency, and four prices with no readable figure were removed rather than re estimated.
- [Four corrections to the board, published together](https://www.agent-verdict.com/pulse/four-corrections-published-together-axes-head-to-head-speed-and-a-withdrawn-duel): The strongest and weakest axes on every tool page were computed over nine axes instead of ten, the head to head W and L record was not actually derived from the duel table, Totalum and Replit's speed scores are corrected against a cited benchmark, and the Lovable versus Totalum duel is withdrawn.
- [The published 0 to 5 rating did not match our own methodology page](https://www.agent-verdict.com/pulse/the-published-star-rating-did-not-match-our-own-methodology): The machine readable star rating on every tool page was read from a stored column following an undocumented curve, not from the community score divided by 200 as the scoring page states. It is corrected today, and the visible rating chip is removed from tool pages.
- [Windsurf has been absorbed into Devin, and leaves the board](https://www.agent-verdict.com/pulse/windsurf-has-been-absorbed-into-devin): windsurf.com now returns an HTTP 308 redirect to devin.ai/desktop. Windsurf's row, votes and head to head record are retired, not transferred: this board does not carry a separate Devin row for them to move to.
- [Gemini CLI's free tier is gone, and its value score reflects that](https://www.agent-verdict.com/pulse/gemini-cli-value-score-cut-after-free-tier-withdrawn): Google shut off free and personal Pro/Ultra access to Gemini CLI on June 18, 2026, replacing it with Antigravity CLI for individuals. The row stays on the board; the value axis and the composite score do not.
- [Mocha has shut down and leaves the board](https://www.agent-verdict.com/pulse/mocha-has-shut-down-and-leaves-the-board): Mocha stopped selling paid plans and confirmed a full shutdown for 1 August 2026. We have removed it rather than keep ranking a product nobody can buy.
- [Marblism did not shut down, it changed category](https://www.agent-verdict.com/pulse/marblism-has-left-the-app-builder-category): Marblism is alive and selling. What it sells now is AI employees, not an app builder. We removed it because the product our tagline described is gone, not because the company is.
- [Hocoos is gone, and its own homepage is why we missed it](https://www.agent-verdict.com/pulse/hocoos-is-gone-and-leaves-the-board): The editor is offline, stored sites were deleted in April 2026, and the pricing page and app subdomain both say so. The marketing homepage kept selling it anyway, which is how it survived our last check.
- [Every tool, every axis, one page: the comparison matrix is live](https://www.agent-verdict.com/pulse/the-comparison-matrix-is-live): All 45 app builders against all 10 scored axes on a single page, winner marked in every column. No more clicking between tool pages to compare an axis across the field.
- [What six points of value actually buys you](https://www.agent-verdict.com/pulse/what-six-points-of-value-actually-buys): Value carries only 6 of the 100 points on this board, and the cheapest tool in a comparison is rarely the one that wins the axis. Here is what the votes actually reward.
- [The app builder board grows from 11 to 45 tracked tools](https://www.agent-verdict.com/pulse/board-expands-from-11-to-45-tracked-app-builders): Every AI app builder we could run the standard brief on now has a row. Nothing about the original 11 got worse this week, several of them just have new company.
- [Framer and Firebase Studio debut straight into S tier](https://www.agent-verdict.com/pulse/framer-and-firebase-studio-debut-in-s-tier): Two of this week's 34 new entries cleared the S tier bar on their first appearance. Neither one is competing for the same job as Lovable or Totalum.
- [Durable takes the small business site duel over Hocoos, 59 to 41](https://www.agent-verdict.com/pulse/durable-wins-the-small-business-site-duel): Both generate a full small business site, copy and logo included, from one prompt. The vote split says the built in CRM and invoicing layer is worth more than a cheaper sticker price.
- [Why our board disagrees with enterprise scorecards, and why that is the point](https://www.agent-verdict.com/pulse/why-our-board-disagrees-with-enterprise-scorecards): Design 22 and speed 21 out of 100. That single decision explains almost every difference between this board and the scorecards written for procurement teams.
- [What agent performance actually measures, and why it is only worth 11 points](https://www.agent-verdict.com/pulse/what-agent-performance-actually-measures): Autonomy is the most impressive axis to demo and the third most important to the people voting. Here is why the community keeps ranking it below design and speed.
- [How the tier bands are earned, and why C tier is not a failing grade](https://www.agent-verdict.com/pulse/how-the-tier-bands-are-earned): S, A, B and C are scope statements as much as quality statements. A tool can be excellent at a small job and sit in C.

## Structured endpoints

- [board.json](https://www.agent-verdict.com/api/board.json): full board as JSON, includes the licence field
- [board.csv](https://www.agent-verdict.com/data/board.csv): full board as CSV
- [llms-full.txt](https://www.agent-verdict.com/llms-full.txt): every tool with strengths, weaknesses, pricing and sources
- [scoring](https://www.agent-verdict.com/scoring): methodology, weights and the seeded vote disclosure
- [movers](https://www.agent-verdict.com/movers): weekly rank deltas, published every Monday
- [matrix](https://www.agent-verdict.com/matrix): all app builders against all 10 scored axes on one page, winner marked per column
- [sitemap index](https://www.agent-verdict.com/sitemap-index.xml)

## Disclosure

Launch scores include editorially seeded votes from our own testing panel. Those are badged Editor panel wherever they appear, and community votes accumulate on top of them.