Browser Use MCP

Browser Use, a widely used open-source library for AI browser agents, ships an MCP server in two forms: a hosted cloud version that runs whole tasks in its own browsers, and a local mode that drives a browser on your computer. In the cloud version Claude describes a task in plain language and follows the agent's progress and results, while the local mode exposes direct navigation, click and extraction tools.

Who it is for

• Teams automating multi-step web tasks
• Developers prototyping browser agents
• Operations staff handling repetitive portal work

Prompts to try

Once it is connected, ask in plain words. Claude decides when to use the server and asks before it changes anything important.

Find the next three public meeting dates on the city council website and list them with their agendas.

Search three job boards for remote product designer roles posted this week and collect the links.

Using my saved profile, open the supplier portal and list last month's invoices.

How to connect it

{
  "mcpServers": {
    "browser-use": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.browser-use.com/v3/mcp",
        "--header",
        "Authorization:${BROWSER_USE_AUTH_HEADER}"
      ],
      "env": {
        "BROWSER_USE_AUTH_HEADER": "Bearer YOUR_API_KEY"
      }
    }
  }
}

How we checked it

Checked Browser Use's cloud and open-source MCP docs, the GitHub repo (MIT, 118k stars, last commit 2026-10-17) and the registry entry com.browser-use/browser-use; the v3 hosted endpoint answers the initialize request.

Source: docs.browser-use.com/cloud/guides/mcp-server