MCP Fetch

Exposes a minimal MCP stdio server with http_request and on-disk caching for chunked large responses.

typescript

GitHub Stars

typescript

Language

6 months ago

First Indexed

3 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "sposer-mcp_fetch": {
      "command": "uvx.exe",
      "args": [
        "--from",
        "E:\\\\Private\\\\Mcp\\\\fetch",
        "mcp-fetch",
        "--transport",
        "stdio"
      ],
      "env": {
        "MCP_FETCH_CACHE_DIR": "./.mcp-fetch-cache",
        "MCP_FETCH_TTL_SECONDS": "1800",
        "MCP_FETCH_MAX_CACHE_BYTES_TOTAL": "536870912",
        "MCP_FETCH_MAX_SINGLE_TRANSFER_BYTES": "209715200",
        "MCP_FETCH_WAIT_CHUNK_TIMEOUT_SECONDS": "30"
      }
    }
  }
}

You run a minimal MCP stdio server called mcp-fetch that exposes a single tool, http_request. It supports all HTTP methods, headers, query parameters, and multiple body types, with chunked responses and on-disk caching so agents can request more data later. This guide shows how to run it, connect with an MCP client, and configure it for your environment.

How to use

Start the MCP server in stdio mode using your MCP runtime tool. You will connect an MCP client to it and issue requests through the http_request tool. The server returns the first chunk of a response along with a transfer_id and offset. If you need more data, request subsequent chunks by providing the transfer_id and the next offset. Use headers to pass HTTP request headers, and rely on chunked transfer for large responses.

You can also run the server in HTTP mode to expose a direct HTTP endpoint locally. In the interactive run, the server will listen at http://127.0.0.1:8000/mcp when started with the HTTP transport. Use MCP_FETCH_TRANSPORT to select the transport and MCP_FETCH_PORT to set the port, or pass --transport http --port 8000 when starting.

How to install

Prerequisites: ensure you have an MCP runtime that supports stdio servers (for example, uvx). You should also have access to a shell or terminal compatible with your operating system.

Install the MCP runtime if you do not already have it.
Download or clone the mcp-fetch component to your machine.
Run the MCP stdio server with the appropriate command and arguments for your environment.

Configuration and notes

The server supports a set of environment variables to tune caching and timeouts. These defaults are used if you do not override them.

Key environment variables and their defaults are as follows:

MCP_FETCH_CACHE_DIR: default ./ .mcp-fetch-cache
MCP_FETCH_TTL_SECONDS: default 1800
MCP_FETCH_MAX_CACHE_BYTES_TOTAL: default 536870912 (512 MiB)
MCP_FETCH_MAX_SINGLE_TRANSFER_BYTES: default 209715200 (200 MiB)
MCP_FETCH_WAIT_CHUNK_TIMEOUT_SECONDS: default 30

Example MCP server configuration

{
  "type": "stdio",
  "name": "mcp_fetch",
  "command": "uvx.exe",
  "args": ["--from", "E:\\Private\\Mcp\\fetch", "mcp-fetch", "--transport", "stdio"]
}

Notes on security and usage

The server is designed to allow private/internal network access by default. Configure your environment to restrict or permit outbound connections as needed for your deployment scenario.

Available tools

http_request

Tool that performs HTTP requests with support for all methods, headers, query params, and chunked responses via a transfer_id mechanism. It handles various body types and caches results on disk for subsequent chunks.

Built by

VeilStrat

AI signals for GTM teams