Skip to content

Stealth browser automation for AI agents over MCP and the Chrome DevTools Protocol: navigation, network hooks, DOM extraction, and pixel-accurate UI cloning.

1.8 KB
Raw Blame

Contributing

Thanks for your interest in improving Stealth Browser MCP. We welcome issues and PRs from the community.

Ways to contribute

  • Report bugs with clear repro steps
  • Propose features with concrete user stories
  • Improve docs and examples
  • Optimize performance or reliability

Development setup

  1. Clone and create a virtualenv
  2. Activate the virtualenv and install deps

Windows

python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt

Mac/Linux

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
  1. Run the server: python src/server.py. Code lives in the stealth_browser_mcp package under src/stealth_browser_mcp/, and pip install -e . also gives you the stealth-browser-mcp command.
  2. Run the tests: python -m unittest discover -s tests -v. Set STEALTH_BROWSER_TESTS=1 to include tests that launch a real browser.
  3. If your change can affect detection, run the stealth benchmark before and after: python -m benchmarks.stealth run --mode headed. See STEALTH_TESTS.md for targets and statuses.

Pull request guidelines

  • Keep PRs focused and under 300 lines of diff when possible
  • Add or update docs when behavior changes
  • Use clear, descriptive titles following Conventional Commits when feasible (e.g., feat: add tab suspend/resume API)
  • Link related issues in the PR description

Issue guidelines

  • For bugs, include expected vs actual behavior, steps to reproduce, logs, and environment details
  • For features, include the problem, target users, and acceptance criteria

Code style

  • Python 3.10+
  • Prefer readable, explicit code and small functions
  • Add minimal docstrings for public functions

Security

If you find a security issue, please follow the process in SECURITY.md.