Cloudflare runs a catalog of official managed remote MCP servers — for docs, Workers builds, observability, DNS analytics, Radar and more — so an AI client can query and help manage your edge. Here's how to connect them.
Setup — official remote servers
Cloudflare's servers are hosted and OAuth-based. Add one as a remote connector — for example the main API/docs server:
{
"mcpServers": {
"cloudflare": {
"url": "https://mcp.cloudflare.com/mcp"
}
}
}
Authorise via Cloudflare OAuth on first use. For stdio-only clients, bridge with npx -y mcp-remote <url>. Cloudflare publishes multiple specialised endpoints (docs, observability, Radar, etc.) — add the ones you need. Restart the client; see MCP config file location.
Pick the right server
It's a catalog, not one server: a docs server (no account needed), an observability server (logs/analytics), a Workers bindings server, DNS analytics, Radar, and more. Add only what you need to keep the tool list lean.
Verify
Ask: "Show error rates for my Worker over the last 24h" (observability) or "How do I add a custom domain to a Worker?" (docs).
Common problems
- Wrong account — re-run OAuth and select the correct Cloudflare account.
- No data — your API token/role lacks access to that product.
mcp-remoteerrors — Node 18+; see Cursor MCP not working.
Going further
Cloudflare pairs with GitHub and Sentry for edge debugging. Browse the devops category or curated loadouts. Hosting your own server? See deploy an MCP server with Docker & Cloudflare.