Check what clients can actually discover
Review whether the remote endpoint exposes clear tools and input schemas that ChatGPT and Claude can scan and understand.
Free MCP publishing preflight
Will your MCP server be ready for ChatGPT and Claude?
A server can work locally and still fail when a real client scans its tools, refreshes authentication, interprets permissions, or tries to recover from an error. Share the remote endpoint and get a practical preflight report before you publish.
The first reports are reviewed manually. You get concrete findings and fixes, not a generic MCP checklist.
ChatGPT + ClaudeTools, auth + safetyFree manual report
What the report covers
From a remote endpoint to a clear launch checklist.
We follow the connection and review path a real MCP client takes, then return the blockers and recommended fixes in priority order.
3 checks
Catch auth and transport problems
Flag remote-connectivity, OAuth, refresh-token, certificate, and Streamable HTTP issues before they slow down a launch.
Make permissions and risk legible
Inspect read, write, and destructive actions for missing annotations, unclear descriptions, and avoidable approval friction.
What the report checks
The details that become launch blockers.
The report follows the path a real client and reviewer must take, from connecting to the endpoint through understanding and safely calling its tools.
We review remote connectivity, transport, authentication, tool discovery, names and descriptions, input schemas, applicable safety annotations, error behavior, and token-efficiency risks.
- Tool discovery and schema clarity
- OAuth, certificates, and remote transport
- Read-only, destructive, and other safety signals
What you receive
A short list of findings you can act on.
You get the blockers first, the evidence behind each finding, and a recommended fix or next test.
The first reports are fulfilled manually so we can distinguish a real launch problem from a theoretical lint warning. The report is independent and does not guarantee approval by OpenAI, Anthropic, or any directory.
Who this is for
You already have a remote MCP endpoint.
This preflight is for developers and product teams preparing to test, publish, or distribute an MCP server.
If you have not built the server yet, use the hosted MCP offer instead. This page intentionally tests the narrower problem of getting an existing server ready for real clients and publishing workflows.
Next step
Find the blockers before the launch
Share the remote MCP endpoint and your email. We will review it manually and send a practical publishing-readiness report for free.