This is deliberately narrow — one publication's back catalogue — and the payoff is that search covers content and tags rather than titles alone, with a relevance score on every hit. The prompt is what shows the intent: it calls the search tool itself to find related posts, so a summary lands in the publication's existing voice rather than a generic one.
A TypeScript MCP server for blog search over a Zenn publication's archive. It also ships an MCP prompt that generates frontend-weekly summaries in that publication's style, calling the search tool itself to find related posts.
- Posts searched across title, description, URL, tags and content, with multiple space-separated keywords treated as OR conditions — `search_cy_fe_articles`
- Results carrying title, publication date, description, link, content, tags and a relevance score, paged by `limit` (default 10, max 100), `offset` and `order`
- A prompt that turns a list of URLs into weekly summaries — a short description, context on the technology, and future implications — `fe-weekly`
- Three ways to run it: `npx @sakupi01/zenn-articles`, the `sakupi/zenn-articles` Docker image, or the hosted endpoint
No account and no key. `npx @sakupi01/zenn-articles` over stdio, or pull the Docker image and run it with `--network=host`. Running the server yourself in code listens on port 8000, with the client pointed at `http://localhost:8000/mcp`; the deployed endpoint is `https://zenn-mcp.sakupi01.com/mcp`. The `query` parameter is required and must be 1 to 100 characters. The package is `@sakupi01/zenn-articles` 1.2.2.
One command — npx @sakupi01/zenn-articles