CSC Documentation MCP Server
Experimental
This service is experimental. Tools and their exact outputs may change, do not rely on it in a production environment.
Model Context Protocol (MCP) is a standardized way for AI agents to access a large variety of tools. See the MCP documentation for a general introduction.
The CSC Documentation MCP server lets agents search the CSC User Guide for up-to-date information about HPC, Roihu, and other CSC services. It indexes the documentation as snippets and returns the snippets that best match a query. The search is semantic, so the agent does not have to guess the exact keywords used in the documentation. A query phrased in the agent's own words is usually enough to find the relevant passages.
The server is public and requires no authentication.
How it works
The server exposes a single tool, search_csc_docs, which takes a search query and
returns the closest matching documentation snippets. The agent decides when to call
it and what to search for.
| Parameter | Description | Values |
|---|---|---|
query |
The search query, in natural language. | Text, up to 100 characters. |
k |
How many of the best matching snippets to return. | An integer from 1 to 10. Default is 4. |
The indexed documentation is kept up-to-date automatically, so the snippets reflect the current CSC User Guide without any action on your part.
The snippets come with links to the corresponding CSC User Guide webpage, so you can ask the agent to cite its sources, and get back links directly to the documentation you're interested in.
Getting started
How you add an MCP server depends on the client, so check the documentation of your agent. Instructions for two common clients are below.
For Claude Code, add it by running:
For OpenCode, run the following command:
If you are adding the server to another client, note that the endpoint is
https://mcp.docs.csc.fi and it uses streamable HTTP.