Langgraph MCP Agents

🚀 MCP Agent Tool Adapter

The MCP Agent Tool Adapter enables modular invocation of tools through the MCP Protocol and provides agents for dynamic inference using Google ADK or LangGraph.

This project transforms MCP tools into:

• 🤖 Google ADK-based agents with streaming FastAPI or CLI interfaces.
• 🧠 LangGraph-based agents using ReAct + streaming tool planning.

🚀 Quick Start

☑️ Prerequisites

• Python 3.10+

⚙️ Installation

# Clone this repository
❯ git clone https://github.com/serkanyasr/mcp-agent-tool-adapter
❯ cd mcp-agent-tool-adapter

# Install dependencies
❯ pip install -r requirements.txt

✨ Features

The MCP Agent Tool Adapter offers the following features:

• Modular tool invocation via the MCP protocol.
• Dynamic inference agents using Google ADK or LangGraph.
• Transformation of MCP tools into different types of agents.

📦 Installation

# Clone this repository
❯ git clone https://github.com/serkanyasr/mcp-agent-tool-adapter
❯ cd mcp-agent-tool-adapter

# Install dependencies
❯ pip install -r requirements.txt

💻 Usage Examples

Basic Usage

Run the Google ADK CLI Agent:

❯ python app_client_adk.py

Run the LangGraph ReAct CLI Agent:

❯ python app_client_langgraph.py

Ensure that your mcp_config.json defines the tools, for example:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["@modular-protocol/filesystem"]
    }
  }
}

📚 Documentation

Project Structure

mcp-agent-tool-adapter/
├── mcp_client/                # Core client implementation (modular)
│   ├── client.py              # MCPClient & MCPServer
│   ├── tool_loader.py         # Advanced asynchronous loader
│   └── types.py               # Shared type definitions
├── app_client_adk.py         # Google ADK agent + CLI chat
├── app_client_langgraph.py   # LangGraph agent + ReAct CLI chat
├── mcp_config.json           # Example MCP tool configuration
├── requirements.txt          # Dependencies
└── README.md

Tool Architecture

MCP tools connect to the agents via MCPClient, handling:

• Tool server.spawn (via subprocess).
• Session management.
• Tool adaptation (langgraph or Google ADK).
• Cleanup work.

You can dynamically switch the agent type by changing tool_type to "google" or "langgraph".

🧠 Technical Details

The MCP Agent Tool Adapter uses the MCP protocol to modularize tool invocation. It leverages Google ADK and LangGraph for dynamic inference, providing different types of agents for various use cases.

📄 License

This project is licensed under the MIT License. For more details, see the LICENSE file.