As a fiber artist, I live on Ravelry–looking up patterns, tracking my projects, managing my queue, logging my stash. I try to keep it as a reference for myself and as a showcase for my crochet projects.

At the same time, I’ve been going deep on AI tooling and looking for a real opportunity to build an MCP server–not a toy, something I’d actually use. The connection was obvious: give Claude Desktop direct access to my Ravelry data, so I could ask about my queue or log a finished project without ever opening a browser.

The Approach

I could have written this myself–TypeScript, REST API, nothing exotic. But I wanted to see how far I could get by letting Claude do the coding while I did the thinking. So far, zero lines written by me.

I fed Claude the Ravelry API documentation and started making decisions out loud: auth model, project structure, which tools to include and which to cut. Claude wrote the code. I reviewed it, caught what broke in production, and directed the next session. Three sessions in, I had something worth shipping.

The result: 21 tools and 5 resources, published on npm as ravelry-mcp. You can run it today with npx ravelry-mcp.

What It Does

The server covers the parts of Ravelry I actually use:

  • Pattern search and details: search by keyword, filter by craft, weight, and category; pull full pattern data
  • Queue management: view your queue, add patterns, remove entries
  • Project tracking: list projects, filter by status (in progress, finished, hibernating, frogged), update project details
  • Library and favorites: browse your saved patterns and stash of saved projects
  • Yarn and stash: search the yarn database, browse your stash entries

Auth is Basic Auth–two env vars in claude_desktop_config.json, one-time setup. No OAuth dance, no hosted relay, no token files.

note
I looked at OAuth 2.0 first–Ravelry supports it. But Ravelry requires https:// redirect URIs, which means a local tool needs a hosted relay just to handle the callback. Once you need hosted infrastructure, the security benefit over a personal API key disappears.

What’s Next

There’s more API surface I haven’t touched yet: individual project pages, photo uploads, full stash CRUD, queue reordering, needle inventory, packs. The core use case works well. I’m hoping to get the fiber artists that also code involved to contribute.

Files & Links

Find it on

← All Projects