Browser Use MCP Server

Browser Use is a browser automation library whose MCP server ships inside the main browser-use Python package rather than as a separate artifact. The codebase carries two stdio server variants. The --mcp flag starts the full-featured server documented on docs.browser-use.com, with tools such as browser_navigate, browser_click, browser_type, browser_get_state, and browser_extract_content, plus an agent tool that requires an LLM API key in the environment. The --cli-mcp flag starts a minimal two-tool server: browser_exec, which runs Python snippets against a browser-harness session, and browser_screenshot.

The distinction matters because the registry entry and the docs point at different variants. The in-repo server.json that feeds the official registry launches the package with --cli-mcp, so a client installing from the registry record gets the minimal exec-style server, not the tool set the public docs walk through. Neither variant carries MCP-level authentication; both run locally.

Browser Use also operates a hosted MCP endpoint at https://api.browser-use.com/v3/mcp, authorized with an x-browser-use-api-key header and exposing cloud session-management tools. That service has a different tool surface from the local server and is not covered by this record.

Authority rests on the domain-verified com.browser-use registry namespace, the GitHub-verified browser-use organization, and the README's registry ownership comment. No surface names a supported MCP specification revision; the library pins the Python mcp==1.26.0 SDK, which is not evidence of revision support. spec_status is therefore unknown. The registry record (0.13.5) trails the repository's latest release (0.13.7, 2026-07-27) by two releases.