Get it free Try it on Telegram

Use Hunch inside Claude and ChatGPT

Hunch is a remote MCP server, so your AI assistant can give it a suspicious message and get a calm second look back: what stands out, and the smallest safe next step. It runs on the same frozen engine as the Hunch extension and site. There is no account and no sign-in, and your message is analyzed, not stored.

What Hunch does inside an assistant

The server URL

Both addresses serve the same server over Streamable HTTP, with no authentication:

Use the branded address where you can. The workers.dev address is a working fallback.

Set up in Claude (step by step)

  1. Open Claude on the web (claude.ai) or the desktop app, signed in to a paid plan.
  2. Go to Settings, then Connectors, and choose Add custom connector.
  3. Give it the name Hunch and paste the server URL above into the URL field.
  4. Leave authentication empty. Hunch needs no sign-in. Save the connector.
  5. In any chat, type something like: Check this message with Hunch, then paste the words you received.

Set up in ChatGPT (step by step)

The official ChatGPT app listing is in review. Today you can connect Hunch through developer mode on an eligible plan:

  1. In ChatGPT, open Settings, then Connectors, and turn on developer mode under the advanced options.
  2. Add a connector and paste the server URL above. Choose the Streamable HTTP transport and no authentication.
  3. Enable the Hunch connector in a chat, then ask it to check a message you are unsure about.

Privacy on this surface

Your message is analyzed, not stored. Hunch reads only the words you paste, never logs or keeps message content, and sends nothing to any outside AI service. There is no account and no sign-in. A fair-use rate limit keeps the free server healthy.

Questions

Can I use Hunch inside Claude or ChatGPT?

Yes. Hunch is a remote MCP server. Add it as a connector in Claude, or in ChatGPT developer mode, using the URL https://mcp.checkyourhunch.com/mcp with no sign-in, then ask your assistant to check a suspicious message.

Do I need an account or an API key?

No. The server is public and read-only, with a fair-use rate limit. There is no account, no API key, and no sign-in.

Does Hunch store the messages I check?

No. Your message is analyzed, not stored. Hunch reads only the words you give it, keeps no message content, and sends nothing to any outside service.

What can Hunch tell me?

For one message it returns a plain finding (stop, worth a check, or looks ordinary), a short reason, and the smallest safe next step, which never reuses a link or phone number from the message. It can also look up whether a channel is one an organization publishes on its own site.

Which assistants are supported?

Any assistant that speaks the Model Context Protocol. Today that includes Claude as a custom connector and ChatGPT in developer mode. The official ChatGPT app listing is in review, and Hunch is published in the official MCP Registry.

Is Hunch a replacement for my own judgment?

No. Hunch gives a calm second look and a safe next step. It never says a message is safe and never calls anyone a scammer. You decide what to do.

Try the free checker on the web →