AutoGen Studio MCP Tools

This repository introduces Model Context Protocol (MCP), a standardized interface for AI agents to discover and utilize various tools, similar to a USB for AI. MCP enables agents to seamlessly integrate new capabilities without needing implementation specifics.

The MCP ecosystem offers diverse servers, including Brave Search for web and local searches, Filesystem for file operations, Playwright for web scraping, MongoDB and SQLite for database interactions, and more. The MCP Servers Repository lists available servers.

To get started, install the mcp package, configure mcp_config.json with server settings (API keys, file paths), and launch AutoGen Studio. MCP uses a dictionary-based approach for dynamic tool usage, allowing agents to list servers, discover tools, and execute them with arguments. Example templates in /gallerys and /teams provide pre-configured setups and workflows.

The client searches for configuration files in the current directory, user config, and environment variables. Development includes running pytest for testing. MCP requires Python 3.12+, Node.js, npx, the MCP Python SDK, and AutoGen Studio.