mcp-supervisor

This repository combines two powerful LangChain projects to create a supervisor-based multi-agent architecture with MCP integration:

LangGraph Supervisor Template

A ready-to-use template for quickly building supervisor-based multi-agent architectures in LangGraph.

LangChain MCP Adapters

A repository that demonstrates how to integrate MCP servers within LangGraph applications.

Overview

This project showcases a nearly no-code approach to building AI assistants.

• Create a supervisor-based multi-agent system with just a few lines of code
• Connect to external services like Zapier MCP (no-code MCP Server management) - https://zapier.com/mcp

Simply include your Zapier MCP server URL in your environment variables, and all these tools become instantly available to your agents

The repository is designed for immediate deployment to LangGraph Cloud using the LangGraph Cloud Quick Start guide. When deployed to LangGraph Cloud, you automatically get LangSmith tracing for comprehensive monitoring and debugging. During development, LangGraph Studio provides a no-code environment for testing and debugging your application.

For a quick and easy user interface, simply connect your application to Agent Chat UI - a pre-built, customizable UI designed specifically for LangGraph applications. This approach eliminates the need to build a frontend from scratch, allowing you to focus on your assistant's capabilities rather than implementation details.

This project leverages the complete LangChain stack for an end-to-end AI application:

Getting Started

1. Clone this repository

   git clone https://github.com/yourusername/mcp-supervisor.git
   cd mcp-supervisor

2. Choose your LLM, agent architecture, and prompts

   • Modify graph.py to select your preferred LLM
   • Customize agent prompts and roles based on your use case
   • Define your tools:
     • Use existing MCP servers (like Zapier)
     • Create custom tools with LangChain

3. Deploy your application

   • Deploy to LangGraph Cloud (recommended) by setting your environment variables as shown in .env.example

4. Use your agent through Agent Chat UI

   • Connect your deployed application to Agent Chat UI for a ready-to-use interface
   • Interact with your multi-agent system through a user-friendly chat interface
   • Test and refine your agent's capabilities