camofox-browser
A headless browser server for AI agents that aims to avoid bot detection, built on Camoufox.
- Stars
- 12.1k
- Forks
- 1.2k
- Licence
- MIT
- Language
- JavaScript
- Last update
- 4 days ago
- Type
- App
What it is
camofox-browser is a server that runs a modified Firefox browser (via the Camoufox project) and exposes it through a REST API so AI agents can load and interact with web pages. The project says its browser changes are made at the C++ level to reduce detection by anti-bot systems. It returns simplified accessibility snapshots instead of raw HTML and gives page elements stable reference IDs for clicking.
- Run an AI agent that opens a page, takes an accessibility snapshot, and clicks an element by its ref
- Use built-in search macros such as @google_search or @youtube_search to run common searches
- Extract YouTube transcripts via the /youtube/transcript endpoint (yt-dlp recommended)
- Log into a site visually through VNC and export storage state for later agent use
Great for
- Developers building AI agents that need to read or interact with websites
- Teams scraping sites that block standard headless Chrome or Playwright
- People who want a self-hosted browser API they can call over HTTP
- Users who want to run the browser on small machines like a Raspberry Pi or low-cost VPS, per the README
Not ideal for
- People who only need a simple, unprotected page fetch; a plain HTTP client is lighter
- Anyone needing guaranteed access to protected sites, since the README claims bypass for some detection but gives no success guarantees
- Users who want a no-code tool; it requires running a Node.js server and calling its API
How to get it
- 1
Clone and build from source
You'll need: Node.js (version not stated in the README) · npm
Clone the repository and enter the folder
git clone https://github.com/jo-inc/camofox-browser && cd camofox-browserInstall dependencies and start the server
npm install && npm start - 2
Run it
- Run the install and start commands above
- Wait for the first-run Camoufox download (about 300MB) to finish
- Open http://localhost:9377 or visit http://localhost:9377/docs for the API docs
- Create a tab and take a snapshot through the REST API
- 3
Get the most out of it
- Use the /docs interactive page to explore all endpoints
- Install yt-dlp locally to speed up the /youtube/transcript endpoint
- Use session isolation to keep cookies and storage separate per user
- Import Netscape-format cookie files for authenticated browsing
- Enable opt-in session tracing to capture Playwright traces for debugging
Downloads
Strengths
- Accessibility snapshots and element refs are designed to keep agent input compact and reliable
- Exposes an OpenAPI spec and interactive docs at /docs
- MIT licensed and actively pushed to, with a large star count
Trade-offs
- Anti-detection claims are the project's own and are not independently verified in the README; results on specific sites will vary
- First run downloads a Camoufox binary of about 300MB, and Docker or Windows setups add extra steps
- Telemetry is enabled by default and sends anonymized crash reports to GitHub Issues; it can be disabled with CAMOFOX_CRASH_REPORT_ENABLED=false
How it compares
| Project | Stars | Type | Level | Updated | Best for |
|---|---|---|---|---|---|
| camofox-browser (this) | 12.1k | App | intermediate | 4 days ago | Developers building AI agents that need to read or interact with websites |
| SeleniumBase | 13.2k | Library | intermediate | 8 days ago | Python developers writing automated browser tests for websites |
| CloakBrowser | 32.7k | Library | beginner | 4 days ago | Python or JavaScript developers already using Playwright who need automation to get past bot checks |
| ego-lite | 21.3k | App | beginner | 4 days ago | Developers who use Claude Code, Codex, or similar agent CLIs and want them to act in a logged-in browser |
| GenericAgent | 14.4k | AI model / agent | intermediate | 4 days ago | Developers who want a small, readable agent codebase they can modify |
| browser-use | 120k | Library | intermediate | 4 days ago | Python developers who want to prototype an AI agent that operates websites |
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.