Cloudflare MCP 410 Gone: Move From /sse to /mcp
A Cloudflare documentation connection that used to work now answers 410 Gone. Is the server gone, and what else changes besides the URL?
- Cloudflare’s retired product-server /sse paths return 410 Gone with a replacement /mcp URL.
- Update that server’s connection, select Streamable HTTP where required, and complete the new sign-in.
- This migration does not retire every MCP server or eliminate SSE as a response format.
Keep the product hostname and change the endpoint
If a Cloudflare MCP connection now returns 410 Gone from a URL ending in /sse, replace it with that server’s /mcp endpoint. The October 22 announcement says the error includes the new URL.
The official server directory lists separate destinations for separate jobs. Migrating a documentation connection does not require replacing it with the broader Cloudflare API server.
| Existing service | Replacement URL |
|---|---|
| Documentation | https://docs.mcp.cloudflare.com/mcp |
| Browser Run | https://browser.mcp.cloudflare.com/mcp |
| Containers | https://containers.mcp.cloudflare.com/mcp |
These are different tools, not interchangeable fallback addresses. Match each configured connection to its original service.
Select Streamable HTTP if the client exposes a transport setting
Cloudflare’s transport documentation identifies Streamable HTTP as the standard remote transport. A client still forced to use legacy SSE needs that setting changed as well as its URL. A client without Streamable HTTP support cannot complete this migration through a URL edit alone.
Seeing SSE in a log is not sufficient evidence of the old transport. The MCP transport specification permits Streamable HTTP to deliver responses using SSE. It replaces the older HTTP+SSE transport with a single endpoint; it does not ban streamed events.
Local stdio connections use a separate mechanism. Leave unrelated local server entries out of this remote-endpoint migration.
A fresh sign-in is part of moving an authenticated connection
Cloudflare says users who signed in through /sse will be asked to sign in again. Reconnect the edited server entry and finish that authorization flow. A login prompt immediately after the change is therefore expected.
Connections already using /mcp are unaffected by this change. Clearing credentials for every MCP server is not a prerequisite for updating one retired endpoint.
Troubleshoot the stage that still fails
Record the effective URL, transport and error code after reconnecting. A retired endpoint, a failed sign-in and an unsuccessful tool call are different stages. If the URL has changed but authorization still fails, the new error needs its own investigation.
There is also a scope distinction for developers hosting their own servers. Cloudflare’s transport guide allows existing McpAgent deployments to retain SSE temporarily during migration. The product-server announcement does not establish that every custom server hosted on Cloudflare has stopped serving it.
Start with the server named in the error, then check its endpoint, transport and authorization in that order. This gives the remaining error a useful context instead of treating every connection failure as the same 410.