Automate your browser with AI.
Website
•
Docs
Browser MCP is an MCP server + Chrome extension that allows you to automate your browser using AI applications like VS Code, Claude, Cursor, and Windsurf.
- ⚡ Fast: Automation happens locally on your machine, resulting in better performance without network latency.
- 🔒 Private: Since automation happens locally, your browser activity stays on your device and isn't sent to remote servers.
- 👤 Logged In: Uses your existing browser profile, keeping you logged into all your services.
- 🥷🏼 Stealth: Avoids basic bot detection and CAPTCHAs by using your real browser fingerprint.
This is a revised version of Browser MCP that fixes the duplicate naming issue in MCP tool commands:
- Fixed: Tool names appearing as
mcp__browsermcp__browser_*(duplicate "browser") - Solution: Changed server name from "Browser MCP" to "browsermcp" to align with package naming conventions
- Result: Tools now appear as
mcp__browsermcp__navigate,mcp__browsermcp__click, etc.
This fix was implemented with assistance from Claude (Anthropic) to improve the MCP integration experience.
This repo contains all the core MCP code for Browser MCP, but currently cannot yet be built on its own due to dependencies on utils and types from the monorepo where it's developed.
Browser MCP was adapted from the Playwright MCP server in order to automate the user's browser rather than creating new browser instances. This allows using the user's existing browser profile to use logged-in sessions and avoid bot detection mechanisms that commonly block automated browser use.