Connects an LLM directly to a real Chrome browser over a single CDP websocket, letting the agent write the helper code and site skills it is missing mid-task.
- Access
- CLI
- Auth
- API Key
- Skills
- 5
Skills
CDP Browser Control
Drives the user running Chrome over a raw CDP websocket to open tabs, click by coordinates, type, and read page state.
Self-Writing Helpers
Writes missing helper functions into agent_helpers.py during execution, so the harness gains new capabilities each run.
Domain Skill Capture
Files site-specific selectors, flows, and edge cases as reusable domain skills in the agent workspace when enabled.
Cloud Browser Sessions
Starts isolated Browser Use Cloud browsers with stealth settings, proxies, and captcha solving for parallel or headless runs.
Session Recordings
Captures local video traces of a browser task between start_recording and stop_recording once recordings are enabled.
Related Agents
agent-browser
CLI · MCPLets coding agents drive Chrome from a native Rust CLI or stdio MCP server, using accessibility snapshots with element…
Playwright MCP
MCP · CLIOfficial Microsoft MCP server for browser automation. Agents navigate pages and interact with elements via…
Moli
CLI · APIFetches, extracts, and automates web pages in an open-source Rust headless browser that outputs Markdown or semantic…
BrowserAct
CLIAutomates real-Chrome browser sessions for AI agents — breaks through anti-bot walls, isolates multi-account sessions…