stagehand

Stagehand MCP Server

A Model Context Protocol (MCP) server that provides AI-powered web automation capabilities using Stagehand. This server enables LLMs to interact with web pages, perform actions, extract data, and observe possible actions in a real browser environment.

Get Started

1. Run npm install to install the necessary dependencies, then run npm run build to get dist/index.js.

2. Set up your Claude Desktop configuration to use the server.

{
  "mcpServers": {
    "stagehand": {
      "command": "node",
      "args": ["path/to/mcp-server-browserbase/stagehand/dist/index.js"],
      "env": {
        "BROWSERBASE_API_KEY": "<YOUR_BROWSERBASE_API_KEY>",
        "BROWSERBASE_PROJECT_ID": "<YOUR_BROWSERBASE_PROJECT_ID>",
        "OPENAI_API_KEY": "<YOUR_OPENAI_API_KEY>",
      }
    }
  }
}

3. Restart your Claude Desktop app and you should see the tools available clicking the 🔨 icon.

4. Start using the tools! Below is a demo video of Claude doing a Google search for OpenAI using stagehand MCP server and Browserbase for a remote headless browser.

Stagehand MCP Server demo - Watch Video

Tools

Stagehand commands

• stagehand_navigate

  • Navigate to any URL in the browser
  • Input:
    • url (string): The URL to navigate to

• stagehand_act

  • Perform an action on the web page
  • Inputs:
    • action (string): The action to perform (e.g., "click the login button")
    • variables (object, optional): Variables used in the action template

• stagehand_extract

  • Extract data from the web page

• stagehand_observe

  • Observe actions that can be performed on the web page
  • Input:
    • instruction (string, optional): Instruction for observation

Resources

The server provides access to one resource:

1. Console Logs (console://logs)

   • Browser console output in text format
   • Includes all console messages from the browser

2. Screenshots (screenshot://<n>)

   • PNG images of captured screenshots
   • Accessible via the screenshot name specified during capture

File Structure

The codebase is organized into the following modules:

• index.ts: Entry point that initializes and runs the server.
• server.ts: Core server logic, including server creation, configuration, and request handling.
• tools.ts: Definitions and implementations of tools that can be called by MCP clients.
• prompts.ts: Prompt templates that can be used by MCP clients.
• resources.ts: Resource definitions and handlers for resource-related requests.
• logging.ts: Comprehensive logging system with rotation and formatting capabilities.
• utils.ts: Utility functions including JSON Schema to Zod schema conversion and message sanitization.

Module Descriptions

index.ts

The main entry point for the application. It:

• Initializes the logging system
• Creates the server instance
• Connects to the stdio transport to receive and respond to requests

server.ts

Contains core server functionality:

• Creates and configures the MCP server
• Defines Stagehand configuration
• Sets up request handlers for all MCP operations
• Manages the Stagehand browser instance

tools.ts

Implements the tools that can be called by MCP clients:

• stagehand_navigate: Navigate to URLs
• stagehand_act: Perform actions on web elements
• stagehand_extract: Extract structured data from web pages
• stagehand_observe: Observe elements on the page
• screenshot: Take screenshots of the current page

prompts.ts

Defines prompt templates for MCP clients:

• click_search_button: Template for clicking search buttons

resources.ts

Manages resources in the MCP protocol:

• Currently provides empty resource and resource template lists

logging.ts

Implements a comprehensive logging system:

• File-based logging with rotation
• In-memory operation logs
• Log formatting and sanitization
• Console logging for debugging

utils.ts

Provides utility functions:

• jsonSchemaToZod: Converts JSON Schema to Zod schema for validation
• sanitizeMessage: Ensures messages are properly formatted JSON

Key Features

• AI-powered web automation
• Perform actions on web pages
• Extract structured data from web pages
• Observe possible actions on web pages
• Simple and extensible API
• Model-agnostic support for various LLM providers

Environment Variables

• BROWSERBASE_API_KEY: API key for BrowserBase authentication
• BROWSERBASE_PROJECT_ID: Project ID for BrowserBase
• OPENAI_API_KEY: API key for OpenAI (used by Stagehand)
• DEBUG: Enable debug logging

MCP Capabilities

This server implements the following MCP capabilities:

• Tools: Allows clients to call tools that control a browser instance
• Prompts: Provides prompt templates for common operations
• Resources: (Currently empty but structured for future expansion)
• Logging: Provides detailed logging capabilities

For more information about the Model Context Protocol, visit:

• MCP Documentation
• MCP Specification

License

Licensed under the MIT License.

Copyright 2024 Browserbase, Inc.