Universal Assistant with LangGraph and Model Context Protocol (MCP)

This project combines LangGraph with the Model Context Protocol (MCP) to build a Universal Assistant. MCP enables seamless integration between LLM applications and external data sources, acting as a standardized interface for connecting AI models to various tools. LangGraph facilitates the creation of complex workflows, representing them as graphs where nodes are actions and edges define information flow.

The assistant uses a multi-agent pattern, routing user messages to the appropriate agent. The agent then selects and invokes the necessary tool on the MCP server. The implementation includes a router that indexes routing information from MCP servers using a vector database, an assistant graph defining node roles and control flow, and a generic MCP wrapper employing a Strategy Pattern for executing operations on MCP servers. This setup allows for flexible and extensible interactions with various MCP-enabled tools and data sources.