How to beginner · 3 min read

How to expose resources in MCP server

Q: How to expose resources in MCP server

To expose resources in an MCP server, define resource handlers and register them with the Server instance from the mcp SDK. Use the stdio_server or other transports to run the server, which listens for requests and exposes your resources to AI agents.

Quick answer

To expose resources in an MCP server, define resource handlers and register them with the Server instance from the mcp SDK. Use the stdio_server or other transports to run the server, which listens for requests and exposes your resources to AI agents.

PREREQUISITES

Python 3.8+
pip install mcp
Basic knowledge of Python async programming

Setup

Install the official mcp Python SDK and prepare your environment.

Ensure Python 3.8 or higher is installed.
Install the MCP SDK via pip.

bash

pip install mcp

Step by step

Create a Python script that defines resource handlers, registers them with an mcp.server.Server instance, and runs the server using stdio_server. This example exposes a simple resource that returns a greeting.

python

import asyncio
from mcp.server import Server
from mcp.server.stdio import stdio_server

async def greet_resource(params):
    name = params.get("name", "world")
    return {"message": f"Hello, {name}!"}

async def main():
    server = Server()
    # Register resource named 'greet'
    server.register_resource("greet", greet_resource)

    # Run the server with stdio transport
    await stdio_server(server)

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

Common variations

You can expose multiple resources by registering more handlers with server.register_resource. For async resource handlers, define them as async functions. MCP supports different transports like stdio_server (default) or SSE-based servers for web integration.

python

import asyncio
from mcp.server import Server
from mcp.server.stdio import stdio_server

async def add_resource(params):
    a = params.get("a", 0)
    b = params.get("b", 0)
    return {"result": a + b}

async def main():
    server = Server()
    server.register_resource("greet", lambda p: {"message": "Hello!"})
    server.register_resource("add", add_resource)
    await stdio_server(server)

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

Troubleshooting

If the server does not start, ensure you are running the script in a terminal that supports stdio communication.
For resource errors, verify your handler functions accept a single params dict and return a dict.
Use asyncio.run() to run the async main function properly.

✅

Key Takeaways

Use the official mcp Python SDK to create and expose resources in an MCP server.
Register resource handlers with Server.register_resource and run the server with stdio_server.
Resource handlers must accept a params dictionary and return a dictionary response.
You can expose multiple resources and use async functions for concurrency.
Troubleshoot by verifying environment compatibility and handler function signatures.

Verified 2026-04

Verify ↗