Fetch MCP Server is a Model Context Protocol server that lets an LLM retrieve a web page and read it as markdown. It reads long pages in chunks and refuses private network addresses by default.
PluginMCP serverFree
Step 1: Install nothing extra if you use uv, or run pip install mcp-server-fetch.
Step 2: Add a fetch entry to your MCP client settings, using uvx with mcp-server-fetch, Docker with mcp/fetch, or python -m mcp_server_fetch.
Step 3: Optionally install Node.js so the server uses a different HTML simplifier.
Step 4: Ask your assistant to fetch a URL and summarize or extract from it.
Step 5: For long pages, have the model call fetch again with a higher start_index.
Not by default. The server refuses private and loopback addresses. Adding the --allow-private-ips argument turns this off, which the README says may be a security risk.
Yes for requests that come from the model. Requests started by the user through a prompt do not check it, and --ignore-robots-txt disables the check entirely.
The fetch tool truncates its output, so call it again with start_index set to where the last chunk ended.
Requirements
Contentrain MCPContentrain MCP is the MCP engine of Contentrain, an open-source content governance toolkit. It lets AI agents extract and update structured content in a Git repository while humans review the changes.
Neruva Agent MCPNeruva is a memory layer for coding agents, delivered as an MCP server backed by a hosted API. Agents recall earlier decisions, log what they do and leave handoff notes between sessions.
KNX Auditor MCPA design-time KNX and ETS6 assistant exposed as an MCP server. It designs, audits and repairs KNX projects and generates Home Assistant configuration, without access to the live bus.