Browserkit

•
1
•

Transforms authenticated user browser sessions into composable, testable AI tools and site-specific MCP servers.

Browserkit is an open-source framework designed to enable AI agents to interact with websites through real, authenticated user browser sessions. It facilitates the creation of site-specific MCP (Multi-Modal Code Interpreter Protocol) servers that run locally on your machine. Each website adapter operates as a dedicated HTTP MCP server with its own headless Chromium instance, allowing multiple AI agents to connect concurrently while protecting the browser session. It supports human-in-the-loop interactions for login and 2FA, and provides advanced features like raw Playwright access via CDP for custom scripting and configurable browser modes.

Key Features

01Local-first and session-persistent operation with headless Chromium
02Raw Playwright access via CDP for external scripting on authenticated sessions
03Auto-registered browser control tools and adapter-specific domain tools
04Human-in-the-loop for login, 2FA, and CAPTCHA handling
050 GitHub stars
06Build site-specific MCP servers over authenticated browser sessions

Use Cases

01Automating complex web tasks for AI tools via programmatic control
02Building AI agents that interact with specific authenticated websites
03Integrating authenticated browser sessions into AI workflows like Cursor or Claude Code