Show HN: Catcher – An AI web testing tool where most tests never hit the API
Catcher is an open-source AI web testing tool that operates locally on users' machines. Unlike traditional AI testing tools that rely on cloud services, Catcher allows users to run tests without sending data off-site. It features a unique vision-coordinate fallback system to enhance test reliability and is designed for ease of use with natural language test descriptions.
- ▪Catcher is a local-first, open-source web testing tool that does not require cloud services.
- ▪Users can describe tests in natural language, and Catcher runs them in a real browser on their machine.
- ▪The tool includes a vision-coordinate fallback feature to improve test accuracy when standard selectors fail.
Opening excerpt (first ~120 words) tap to expand
Catcher English | 简体中文 Open-source, local-first, BYOK AI web testing. Describe tests in English, run them in a real browser on your machine. ✨ How it's different Most AI testing tools are paid SaaS that runs your tests on their cloud with their LLM. Catcher is the opposite: Desktop app, not a service — your sites, sessions, cookies and screenshots never leave your machine BYOK LLM — point it at OpenAI / Anthropic / Gemini / Ollama / any OpenAI-compatible endpoint; you pay the provider directly Vision-coordinate fallback — when a click misses through every selector strategy, Catcher screenshots the page and asks the LLM to point at {x, y}.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at GitHub.