Media
2 items
Description
Connect your Chrome browser to an upstream you trust for explicitly authorized MCP browser automation. Connect the Chrome browser you control to trusted AI tools over the Model Context Protocol (MCP). Browser MCP bridges your browser to either the managed Browser MCP service or a self-hosted upstream server. Sign in with GitHub and explicitly approve the browser connection, or enter your own upstream URL and browser connection token. Generate a separate, browser-scoped MCP token in the dashboard and use it with a compatible Streamable HTTP MCP client. Tools include: • List, open, navigate and close HTTP(S) tabs. • Read bounded page snapshots and interactive element selectors. • Click, type, press keys and scroll. • Capture screenshots, wait for page conditions and run JavaScript in a page through Chrome's debugger API. You remain in control: explicitly consent before connecting, pause or disconnect in the extension, and revoke client credentials in the dashboard. Chrome's debugger indicator remains visible during control. The upstream control-plane origin and privileged Chrome pages are excluded. IMPORTANT: MCP access grants broad browser control, including signed-in websites, page contents and arbitrary page JavaScript. It is not read-only or limited to one tab. Use a separate Chrome profile and connect only servers, AI clients and models you trust. Browser data and screenshots pass through your selected upstream server to your MCP client; this is not end-to-end encryption that hides results from the server operator. Your AI provider may have its own retention policies. Chrome 120 or later is required. A managed account or self-hosted Browser MCP server and a compatible MCP client are required for remote automation. Self-hosted community servers support configurable OIDC and token-based pairing. Setup and tool reference: https://browser-mcp.mintlify.site Privacy and security: https://browser-mcp.mintlify.site/security