MCP To Langchain Addapter

🚀 MCP to LangChain and LangGraph Adapter Documentation

This adapter documentation provides a comprehensive guide on integrating MCP tools with LangChain and LangGraph, offering seamless tool adaptation and management.

🚀 Quick Start

Initialize the Adapter

from mcp_langchain_adapter import MCPAdapter

# Create an adapter instance and specify the MCP server script path
adapter = MCPAdapter(server_script_path="path/to/your/mcp_server.py")

# Synchronously initialize the connection to the MCP server
adapter.initialize()

Get the Tool List

tools = adapter.get_tools()

# Print information about available tools
print(f"Found {len(tools)} tools:")
for tool in tools:
    print(f"- {tool.name}: {tool.description}")

Tool Usage Example

# Get a tool by a specific name
add_tool = adapter.get_tool_by_name("add")

if add_tool:
    # Use the tool to perform an operation
    result = add_tool.run({"a": 5, "b": 7})
    print(f"Addition result: {result}")

weather_tool = adapter.get_tool_by_name("get_weather")
if weather_tool:
    result = weather_tool.run({"city": "London"})
    print(f"Weather query result: {result}")

📦 Installation Guide

Install Dependencies

pip install mcp-langchain-adapter

✨ Features

Tool Adaptation

• Convert MCP protocol tools into the BaseTool class compatible with LangChain and LangGraph.
• Support custom parameter passing, including environment variables.

Server Startup and Management

• Provide a simple method to start the MCP server process.
• Support asynchronous initialization and resource cleanup.

💻 Usage Examples

Basic Usage

from mcp_langchain_adapter import MCPAdapter

adapter = MCPAdapter("path/to/mcp_server.py")
adapter.initialize()

# Get all available tools
tools = adapter.get_tools()
print(f"Detected {len(tools)} tools:")

for tool in tools:
    print(f"- Name: {tool.name}")
    print(f"  Description: {tool.description}")
    print(f"  Parameters: {tool.args_schema}\n")

Advanced Usage

from mcp_langchain_adapter import MCPToolWrapper

# Create a tool wrapper instance
wrapper = MCPToolWrapper(
    name="Calculator",
    description="A tool for performing basic arithmetic operations",
    server_script_path="path/to/calculator_server.py",
    env={"ENV_VAR": "value"},
    args_schema={  # Optional parameter definition
        "a": {"type": "int", "required": True},
        "b": {"type": "int", "required": True}
    }
)

Automatic Detection and Discovery

• Automatically discover MCP tools by scanning a specified directory.
• Support dynamic loading of newly added tools.

📚 Documentation

Troubleshooting

Common Issues

1. Unable to connect to the MCP server?

   • Check if the server script path is correct.
   • Ensure that the server is running and listening on the correct port.

2. Tools not found?

   • Confirm that the tools meet the adapter's specifications.
   • Check the log output for more information.

3. Incorrect environment variable configuration?

   • Check if the env parameter is passed correctly.
   • Ensure that the server can read these environment variables.

Logging

# Configure logging
import logging
logging.basicConfig(level=logging.DEBUG)

adapter = MCPAdapter("path/to/mcp_server.py", log_level=logging.DEBUG)

Custom Extension

Custom Tool Loading

def custom_tool_loader(directory):
    # Manually scan the specified directory and return a list of MCP tools
    tools = []
    for script in os.listdir(directory):
        if script.endswith(".py"):
            try:
                tool = MCPToolWrapper(..., server_script_path=os.path.join(directory, script))
                tools.append(tool)
            except Exception as e:
                print(f"Error loading script {script}: {e}")
    return tools

adapter = MCPAdapter(custom_tool_loader=custom_tool_loader)

Asynchronous Support

async def async_usage():
    adapter = MCPAdapter("path/to/mcp_server.py")
    await adapter.initialize()  # Asynchronously initialize the connection

    # Example of asynchronous tool call
    result = await add_tool.async_run({"a": 5, "b": 7})
    print(f"Asynchronous addition result: {result}")

More Resources

• MCP Protocol Official Documentation
• LangChain Community Support

Contribution Guide

Contributions to bug fixes and feature enhancements are welcome! Please fork this repository and read Contributing.md before submitting your changes.

📄 License

MIT License

Contact Information

• Email: support@mcp-langchain-adapter.com
• GitHub: https://github.com/mcp/langchain-adapter