Browserbase MCP Server

Example Use Cases

• 1

  Programmatically navigate websites, interact with UI elements, and perform form submissions or clicks to complete web workflows.

• 2

  High-fidelity visual context—capture full-page or element screenshots (with annotations) to support vision-enabled reasoning or data extraction.

• 3

  Extract structured information from web pages, maintain session state across interactions, and use proxies or stealth features for robust scraping and automation.

Description

This MCP server enables LLMs to control cloud browsers to navigate pages, click, fill forms, take full-page or element screenshots, and extract structured data with precise, automatable actions. It supports multiple LLM providers (Gemini by default, plus OpenAI, Claude, and custom models via Stagehand), session persistence, proxy and stealth options, and vision-enabled annotated screenshots. Deployable locally (STDIO/Docker) or via a hosted SHTTP endpoint (smithery.ai) and configurable with CLI flags for proxies, viewport sizing, context persistence, and custom models.