freetoken.link
Menu
Sign in
Compatibility guide

“OpenAI-compatible” does not always mean Codex-ready.

Many free inference providers implement Chat Completions. Codex may expect the Responses API, so the safe workflow checks that boundary before treating a copied config as working.

Key ruleKeep credentials in environment variables

Never embed a real key in a shareable config URL or screenshot.

Start with the protocol, not the model name

Confirm whether the provider documents the endpoint your client actually uses. A successful /chat/completions call does not prove /responses support.

Use the smallest safe test

  1. Verify the key against an allowlisted provider endpoint.
  2. Run one prompt in the browser playground if CORS permits it.
  3. Generate a config that keeps the key in an environment variable.
  4. If Codex rejects the wire format, use a trusted local compatibility adapter instead of sending the key to a web proxy.
The FreeToken config generator labels the compatibility caveat; it does not claim every OpenAI-shaped endpoint supports Codex.
Open tool setup