Model integrity

Verify what actually ran

Token Harbor never silently swaps your model for a cheaper one. Every response tells you the real upstream model that served it — and you can prove it here.

How to verify

  1. Make any API call. The JSON response has an id field like chatcmpl-…. It’s also returned in the X-TH-Request-Id response header.
  2. Paste that id above (or open the link you were sent).
  3. See the exact upstream model, tokens and latency for that request. Programmatic check: GET /api/verify/<id>.

Only routing metadata is shown — never your prompt, response, identity or cost.