briqt/my-agent-browser

my-agent-browser is a thin, config-driven wrapper that exposes native MCP tool calls for browser automation (navigate_page, take_snapshot, click, fill, evaluate_script, etc.) so AI agents can control Chrome without custom runtime code. It's intended for agents that support MCP (examples: Claude Code, Cursor, Kiro). Key features: - start-mcp.js: cross-platform MCP server wrapper that launches and manages a detached Chrome instance. - Shared Chrome process with browser.lock to track clients; last client shuts down Chrome. - All settings in ~/.config/agent-skills/my-agent-browser/config.json with JSON Schema validation. - Documentation for setup, troubleshooting, advanced tooling (Lighthouse, console, emulation), network debugging, scraping patterns and multi-tab workflows. Installation via npx skills add and chrome-devtools-mcp. Configuration changes apply on next session.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.