Cloudflare's documentation server lets your AI agent search the current Cloudflare
docs before it writes code for Workers, R2, D1 or anything else on the platform.
Cloudflare ships changes often, and a model's memory of a wrangler config or a
binding's API goes stale fast. It is a hosted server with two read-only tools and
no sign-in, which makes it a low-risk first MCP server to try. This site
runs on Cloudflare Workers, which is why it made our list.
Prompts to try
- "How do I bind a D1 database to a Worker in wrangler.jsonc? Check the Cloudflare docs."
- "What are the limits on Workers CPU time on the free plan?"
- "I have a Pages project with functions. What do I need to change to move it to Workers?"
- "Show me how to cache a fetch response in a Worker with the Cache API."
Gotchas
- It returns passages, not whole pages. Results are chunks of the docs that match the query. For a long guide, ask it to search again with a narrower query.
- Your queries go to Cloudflare. Nothing else does: no code, no files. There is no account, so nothing ties a query to you beyond your connection.
- It is one of many. Cloudflare runs a server per product area. This page is only the docs one; most of the others sign in to your Cloudflare account, and some of them can change things there.
When not to use it
If you do not build on Cloudflare, skip it. For other libraries' docs, try the Context7 MCP server.