MCP Server Fetch

2025.4.7 · active · verified Mon Apr 13

MCP Server Fetch is a Python library that provides a server implementation for the Model Context Protocol (MCP). It's designed to fetch, process, and convert web content into a format suitable for use by Large Language Models (LLMs). As of version `2025.4.7`, it offers a robust solution for web data ingestion within an LLM ecosystem. The project follows a rapid, approximately monthly, release cadence, often aligning with a YYYY.M.D versioning scheme.

Warnings

gotcha It is crucial to understand that `mcp-server-fetch` provides the server application itself, while programmatic interaction with a running server (e.g., fetching URLs) requires the separate `mcp-client` library. Do not attempt to import client-side utilities like `MCPClient` from `mcp-server-fetch`.
Fix: For client-side operations, `pip install mcp-client` and import from `mcp_client`.
gotcha The server's behavior and accessibility (e.g., CORS, debug settings, API keys) are heavily reliant on environment variables. Failing to set these correctly can lead to unexpected errors or security vulnerabilities.
Fix: Consult the official documentation for a comprehensive list of environment variables and set them appropriately before starting the server. Examples include `CORS_ORIGINS`, `MCP_DEBUG`, etc.
breaking Due to its rapid development cycle (monthly releases) and its position within an evolving protocol ecosystem, API stability between major monthly versions (e.g., from `2025.3.x` to `2025.4.x`) may not be strictly guaranteed. Breaking changes, while not always explicitly detailed in changelogs, can occur.
Fix: Pin your dependency to a specific `YYYY.M.patch` version (e.g., `mcp-server-fetch==2025.4.7`) and thoroughly test when upgrading across month boundaries.
gotcha Running the server programmatically as shown in the quickstart requires `uvicorn` to be installed. While `mcp-server-fetch` lists `uvicorn` as an optional dependency (via `[all]` extra), it's a runtime necessity for most deployments.
Fix: Ensure `uvicorn` is installed in your environment, either directly (`pip install uvicorn`) or by using the `[all]` extra during `mcp-server-fetch` installation (`pip install 'mcp-server-fetch[all]'`).

Install

pip install mcp-server-fetch Install core library
pip install 'mcp-server-fetch[all]' Install with all optional dependencies

Imports

create_app
```
from mcp_server_fetch.server import create_app
```
MCPClient is part of the separate `mcp-client` library, which is used to interact with a running `mcp-server-fetch` instance. The `mcp-server-fetch` library itself primarily provides the server application, accessible via `create_app` for programmatic setup.

Quickstart

This quickstart demonstrates how to programmatically start the MCP Server Fetch application using `uvicorn`. It first configures essential environment variables (like CORS) which are critical for its operation, then creates the FastAPI app instance, and finally runs it using `uvicorn.run`. For interaction with the running server, install and use the separate `mcp-client` library.

import uvicorn
from mcp_server_fetch.server import create_app
import os
import asyncio

# Configure necessary environment variables, e.g., for CORS or debug mode
os.environ['CORS_ORIGINS'] = 'http://localhost:3000,http://127.0.0.1:3000'
os.environ['MCP_DEBUG'] = 'true'

# Create the FastAPI application instance
app = create_app()

async def run_server():
    print("Starting MCP Server Fetch on http://127.0.0.1:8000")
    print("Use Ctrl+C to stop the server.")
    config = uvicorn.Config(app, host="127.0.0.1", port=8000, log_level="info")
    server = uvicorn.Server(config)
    await server.serve()

if __name__ == "__main__":
    # This demonstrates programmatic startup. More commonly, you'd run from CLI:
    # uvicorn mcp_server_fetch.server:app --host 127.0.0.1 --port 8000
    try:
        asyncio.run(run_server())
    except KeyboardInterrupt:
        print("\nServer stopped.")

view raw JSON →