Browserbase MCP Server
Browserbase MCP Server gives agents cloud browser control through Browserbase and Stagehand, with both hosted HTTP and self-hosted stdio options. It suits teams that want MCP-based browser automation but prefer Browserbase sessions and Stagehand’s action model instead of running Playwright locally.
npx skills add agentskillexchange/skills --skill browserbase-mcp-server
Browserbase MCP Server is Browserbase’s MCP implementation for cloud browser automation. It connects an MCP client to Browserbase sessions and Stagehand so an agent can start a browser session, navigate pages, observe actionable elements, execute actions, and extract information from live sites. The upstream README documents both a hosted MCP endpoint and a self-hosted npm package, which makes it flexible for teams that want a quick managed setup or tighter control over deployment.
What this skill actually does
The project exposes a focused set of browser tools including start, end, navigate, act, observe, and extract. Browserbase positions this as a standardized MCP bridge for LLM applications, while Stagehand provides the browser interaction layer and default model-backed action engine. For self-hosted operation, the maintainer documents an npx @browserbasehq/mcp flow with environment variables for BROWSERBASE_API_KEY, BROWSERBASE_PROJECT_ID, and optionally GEMINI_API_KEY.
This belongs in Browser Automation because the real job-to-be-done is running web actions in a browser session from an MCP client, not just scraping static HTML. It also clearly maps to the MCP framework and has strong external verification signals: an official GitHub repository, npm package, published documentation, open-source license, and healthy adoption. For users already invested in Browserbase or Stagehand, this is one of the most direct ways to expose those capabilities to agentic tools.