wouldit.work

AI tools

Screenshot to Code: an 81K-star GitHub repo that became a $263K bootstrapped app, now for sale

Abi Raja's open-source screenshot-to-code tool turned GitHub stars into a hosted product with Stripe-verified revenue. Now he has a new job and has listed it for a quick sale. Here is what the numbers on TrustMRR and GitHub show.

Founder
Abi Raja
Launched
2023
Model
Open-source repo plus a hosted app with credit subscriptions (Hobby $15/month, Pro $40/month)

Published · 2 min read · 3 sources

Key takeaways

  • The MIT-licensed GitHub repo, created in Nov 2023, has about 81K stars and 9.9K forks (checked 4 Nov 2026).
  • TrustMRR (Stripe-verified, 4 Nov 2026): $262,731 all-time revenue, $4,460 MRR, 294 active subscriptions, bootstrapped, 1 person.
  • The founder listed it for sale because he is starting a full-time job; asking price $18,000, cut from $38,000.

What it is

Screenshot to Code turns a screenshot, mockup, Figma design or screen recording into front-end code such as HTML/Tailwind, React or Vue (GitHub README). The code is MIT-licensed on GitHub, and the README points people to the official hosted product at screenshottocode.com (GitHub). Running it yourself needs your own OpenAI, Anthropic or Gemini API key (GitHub README).

Timeline

  • 14 Nov 2023: the GitHub repository abi/screenshot-to-code is created (GitHub); TrustMRR lists the business as founded in November 2023 (TrustMRR).
  • Mar 2024: Abi Raja publishes an evaluation of Claude 3 against GPT-4 vision on 16 screenshots, calling the repo "a very popular open source project" (blog post in the repo).
  • Oct 2026: the hosted business goes up for sale on TrustMRR. The asking price went from $38,000 to $18,000 (TrustMRR).
  • 4 Nov 2026: about 81K stars and 9.9K forks on GitHub (GitHub); $262,731 all-time revenue, $4,460 MRR and 294 active subscriptions, verified with a Stripe API key (TrustMRR).

Why it worked

  • The repo is the marketing. In his sale note the founder says the 80K-star open-source repo "drives steady organic traffic" (TrustMRR), and the README's first link is to the hosted app (GitHub).
  • Self-hosting has friction on purpose. Running it locally means API keys plus a React/Vite frontend and a FastAPI backend (GitHub README), so many people pay for the hosted version instead.
  • Riding each new model. He tested new vision models as they came out, such as Claude 3 in March 2024 (blog post), and the README now lists several current model families (GitHub).
  • Low upkeep. The founder says it takes "under 2 hrs/week to run" (TrustMRR).

What to learn

  1. A popular open-source repo can be a free acquisition channel for a paid hosted version, as long as the hosted version saves real setup work.
  2. Stars and registered users are not paying users. Here, 258K registered users became about 300 subscribers, by the founder's own count (TrustMRR).
  3. AI wrappers carry model costs. Watch margin as closely as revenue.

Risks and what didn't work

  • MRR is small next to the reach: $4,460 MRR against about 81K stars (TrustMRR, GitHub).
  • Model costs bite. The founder reports a 12-month margin of 60% and hopes a "recent switch to cheaper AI models" lifts it (TrustMRR).
  • Anyone can self-host the MIT code, and AI coding tools now generate UIs from images too, so the hosted app has to stay easier than the alternatives.
  • The founder is leaving to start "a new full-time job" (TrustMRR); the user and margin figures in his note are self-reported.

FAQ

Short answers drawn from the sections above.

What is Screenshot to Code?

Screenshot to Code turns a screenshot, mockup, Figma design or screen recording into front-end code such as HTML/Tailwind, React or Vue (GitHub README). The code is MIT-licensed on GitHub, and the README points people to the official hosted product at screenshottocode.com (GitHub). Running it yourself needs your own OpenAI, Anthropic or Gemini API key (GitHub README).

Why did Screenshot to Code work?

  • The repo is the marketing. In his sale note the founder says the 80K-star open-source repo "drives steady organic traffic" (TrustMRR), and the README's first link is to the hosted app (GitHub).
  • Self-hosting has friction on purpose. Running it locally means API keys plus a React/Vite frontend and a FastAPI backend (GitHub README), so many people pay for the hosted version instead.
  • Riding each new model. He tested new vision models as they came out, such as Claude 3 in March 2024 (blog post), and the README now lists several current model families (GitHub).
  • Low upkeep. The founder says it takes "under 2 hrs/week to run" (TrustMRR).

What can indie builders learn from Screenshot to Code?

  1. A popular open-source repo can be a free acquisition channel for a paid hosted version, as long as the hosted version saves real setup work.
  2. Stars and registered users are not paying users. Here, 258K registered users became about 300 subscribers, by the founder's own count (TrustMRR).
  3. AI wrappers carry model costs. Watch margin as closely as revenue.

What didn't work for Screenshot to Code, and what are the risks?

  • MRR is small next to the reach: $4,460 MRR against about 81K stars (TrustMRR, GitHub).
  • Model costs bite. The founder reports a 12-month margin of 60% and hopes a "recent switch to cheaper AI models" lifts it (TrustMRR).
  • Anyone can self-host the MIT code, and AI coding tools now generate UIs from images too, so the hosted app has to stay easier than the alternatives.
  • The founder is leaving to start "a new full-time job" (TrustMRR); the user and margin figures in his note are self-reported.

Sources

Every link cited above, in order of first use.

  1. 1.GitHub README (github.com)
  2. 2.TrustMRR (trustmrr.com)
  3. 3.blog post in the repo (github.com)

Related case studies

More in AI tools →