Summary
PipeCD's docs become something the agent can quote.
Deployment questions get answered out of the project's own documentation rather than from whatever the model remembers. The search narrows to the right page and the read brings the page itself back, so an answer arrives with the text it came from.
What it is
A local MCP server over the official PipeCD documentation. It clones the docs from GitHub into a temporary directory, indexes the Markdown by title and content, and answers full-text searches against that index.
What you get
- Full-text search across the PipeCD docs, with space-separated keywords treated as AND
- Results paged with an offset and a limit, twenty by default
- The full content of a page returned by its path
Requirements
Npx, plus a GitHub personal access token with read:packages scope — the package is published to GitHub Packages, so the registry line and the token go in your .npmrc first.