About
An MCP server that lets AI agents directly control a local browser. Supports page navigation, content extraction, screenshots, form filling, clicking, and more — letting agents browse the web like a human.
Our Verdict
Browser MCP brings browser capabilities to the MCP ecosystem. Its GitHub popularity reflects community trust, and the standardized protocol ensures broad compatibility across AI clients.
Pricing
Use Cases
- Integrating browser capabilities into AI agent workflows
- Building custom AI-powered tools with Browser MCP
- Enhancing AI agent functionality with standardized MCP access
Pros
- Browser MCP provides standardized MCP protocol access
- Open source with 6,912 GitHub stars
- Active development and community support
Cons
- Requires MCP-compatible client (Claude, Cursor, etc.)
- Setup complexity varies by client
- Some features may depend on client capabilities
Installation
npx -y @browsermcp/mcpFrequently Asked Questions
How is Browser MCP different from Playwright MCP?
Browser MCP controls your local real browser (Chrome/Firefox, etc.) that you're already using — it can access logged-in websites, existing sessions, and cookies. Playwright MCP launches a headless browser, making it better suited for automated testing and crawling. They complement each other.
What automation tasks can Browser MCP do?
It can auto-fill and submit forms, extract web content, take screenshots, click buttons and links, scroll pages, manage multiple tabs, and read login states. Suitable for data collection, page monitoring, automated testing, and content management.
Is Browser MCP safe? Will it leak my browsing data?
The MCP server runs locally and communicates with your browser via the Chrome DevTools Protocol (CDP). All data is processed locally and never uploaded to third parties. AI agents only operate the browser when explicitly instructed. Review the sites your agent is asked to visit.