Model Context Protocol (MCP)

Learn how to use MCP with Agno to enable your agents to interact with external systems through a standardized interface.

The Model Context Protocol (MCP) enables Agents to interact with external systems through a standardized interface. You can connect your Agents to any MCP server, using Agno’s MCP integration.

Usage

1. Find the MCP server you want to use

   You can use any working MCP server. To see some examples, you can check this GitHub repository, by the maintainers of the MCP themselves.

2. Initialize the MCP integration

   Intialize the MCPTools class as a context manager. The recommended way to define the MCP server, is to use the command or url parameters. With command, you can pass the command used to run the MCP server you want. With url, you can pass the URL of the running MCP server you want to use.

   For example, to use the “mcp-server-git” server, you can do the following:

   from agno.tools.mcp import MCPTools
   
   async with MCPTools(command=f"uvx mcp-server-git") as mcp_tools:
       ...
   

3. Provide the MCPTools to the Agent

   When initializing the Agent, pass the MCPTools class in the tools parameter.

   The agent will now be ready to use the MCP server:

   from agno.agent import Agent
   from agno.models.openai import OpenAIChat
   from agno.tools.mcp import MCPTools
   
   async with MCPTools(command=f"uvx mcp-server-git") as mcp_tools:
       # Setup and run the agent
       agent = Agent(model=OpenAIChat(id="gpt-4o"), tools=[mcp_tools])
       await agent.aprint_response("What is the license for this project?", stream=True)
   

Basic example: Filesystem Agent

Here’s a filesystem agent that uses the Filesystem MCP server to explore and analyze files:

filesystem_agent.py

import asyncio
from pathlib import Path
from textwrap import dedent

from agno.agent import Agent
from agno.models.openai import OpenAIChat
from agno.tools.mcp import MCPTools
from mcp import StdioServerParameters


async def run_agent(message: str) -> None:
    """Run the filesystem agent with the given message."""

    file_path = str(Path(__file__).parent.parent.parent.parent)

    # MCP server to access the filesystem (via `npx`)
    async with MCPTools(f"npx -y @modelcontextprotocol/server-filesystem {file_path}") as mcp_tools:
        agent = Agent(
            model=OpenAIChat(id="gpt-4o"),
            tools=[mcp_tools],
            instructions=dedent("""\
                You are a filesystem assistant. Help users explore files and directories.

                - Navigate the filesystem to answer questions
                - Use the list_allowed_directories tool to find directories that you can access
                - Provide clear context about files you examine
                - Use headings to organize your responses
                - Be concise and focus on relevant information\
            """),
            markdown=True,
            show_tool_calls=True,
        )

        # Run the agent
        await agent.aprint_response(message, stream=True)


# Example usage
if __name__ == "__main__":
    # Basic example - exploring project license
    asyncio.run(run_agent("What is the license for this project?"))

Using MCP in Agno Playground

You can also run MCP servers in the Agno Playground, which provides a web interface for interacting with your agents. Here’s an example of a GitHub agent running in the Playground:

github_playground.py

import asyncio
from os import getenv
from textwrap import dedent

import nest_asyncio
from agno.agent import Agent
from agno.models.openai import OpenAIChat
from agno.playground import Playground
from agno.storage.agent.sqlite import SqliteAgentStorage
from agno.tools.mcp import MCPTools

# Allow nested event loops
nest_asyncio.apply()

agent_storage_file: str = "tmp/agents.db"


async def run_server() -> None:
    """Run the GitHub agent server."""
    github_token = getenv("GITHUB_TOKEN") or getenv("GITHUB_ACCESS_TOKEN")
    if not github_token:
        raise ValueError("GITHUB_TOKEN environment variable is required")

    # Create a client session to connect to the MCP server
    async with MCPTools("npx -y @modelcontextprotocol/server-github") as mcp_tools:
        agent = Agent(
            name="MCP GitHub Agent",
            tools=[mcp_tools],
            instructions=dedent("""\
                You are a GitHub assistant. Help users explore repositories and their activity.

                - Use headings to organize your responses
                - Be concise and focus on relevant information\
            """),
            model=OpenAIChat(id="gpt-4o"),
            storage=SqliteAgentStorage(
                table_name="basic_agent",
                db_file=agent_storage_file,
                auto_upgrade_schema=True,
            ),
            add_history_to_messages=True,
            num_history_responses=3,
            add_datetime_to_instructions=True,
            markdown=True,
        )

        playground = Playground(agents=[agent])
        app = playground.get_app()

        # Serve the app while keeping the MCPTools context manager alive
        playground.serve(app)


if __name__ == "__main__":
    asyncio.run(run_server())

Best Practices

1. Error Handling: Always include proper error handling for MCP server connections and operations.

2. Resource Cleanup: Use MCPTools or MultiMCPTools as an async context manager to ensure proper cleanup of resources:

async with MCPTools(command) as mcp_tools:
    # Your agent code here

3. Clear Instructions: Provide clear and specific instructions to your agent:

instructions = """
You are a filesystem assistant. Help users explore files and directories.
- Navigate the filesystem to answer questions
- Use the list_allowed_directories tool to find accessible directories
- Provide clear context about files you examine
- Be concise and focus on relevant information
"""

More Information

• Find examples of Agents that use MCP here.
• Find a collection of MCP servers here.
• Read the MCP documentation to learn more about the Model Context Protocol.
• Checkout the Agno Cookbook for more examples of Agents that use MCP.

Assistant

Responses are generated using AI and may contain mistakes.