Browser Use

Browser Use's own MCP server lets your agent hand web tasks to Browser Use. Describe a task in plain words and a Browser Use agent runs it in a stealth cloud browser: it moves through sites, fills in forms and pulls out the data you asked for, then reports back. What you can do with it - Run a web task in a cloud browser session and get the result back in the chat - Ask for the result in a shape you define, with an output schema - Keep a session open and send it follow-up tasks - Check a session's status, step count, cost and live view link - Read the agent's messages: its browser actions, reasoning and results - Use saved browser profiles for tasks on sites where you're signed in - List recent sessions, and stop a task or the whole session How it connects Browser Use hosts it at https://api.browser-use.com/v3/mcp. Add your Browser Use Cloud API key, from cloud.browser-use.com/settings, as a bearer token in the Authorization header (the older X-Browser-Use-API-Key header still works). It runs Browser Use Cloud V3 sessions. Browser Use's docs show setup for Claude Code, Claude Desktop, Cursor and Windsurf. You can also run it on your own machine from the open-source browser-use Python package with uvx. The local server gives your agent direct control of a browser. Good to know - Browser Use Cloud is pay as you go, so cloud sessions use credits on your account. - Sessions use a US proxy by default; you can pick another country or turn the proxy off. - Keep your API key secret, and don't commit it to a repository. - The local server can reach your browser and file system, so Browser Use advises connecting only clients you trust.