Guide EN-02
Codex API Proxy and Relay Providers
For Codex and Codex CLI users, review OpenAI-compatible endpoints, streaming behavior, model availability, and error rates before daily use.
Good fit
Personal projects and low-risk Codex CLI use.
Check first
Base URL, model names, streaming, JSON mode, and rate limits.
Main risk
Model substitution, high error rates, and unclear prompt logging.
Comparison
For Codex, compatibility is more important than headline pricing.
| Provider | OpenAI compatibility | Setup | Risk | Updated |
|---|---|---|---|---|
| QuickRouter | OpenAI-compatible API | Base URL pending test | Medium risk | 2026-05-18 |
| PatewayAI | Codex tool adaptation | Base URL pending test | Medium risk | 2026-05-18 |
| OhMyGPT | Explicit Codex support | Base URL pending test | Medium risk | 2026-05-18 |
| Kokomi API | Codex / Claude Code focused | Base URL pending test | Needs review | 2026-05-18 |
Setup
Use a small test before pointing Codex at a relay.
- Confirm the provider supports your target Codex or OpenAI-compatible model.
- Set the base URL and API key in your tool environment.
- Record errors, latency, and model behavior.
OPENAI_BASE_URL=https://example.com/v1Risk boundary
Code privacy
Avoid unknown relays for company repositories.
Context length
Test long-context coding tasks.
Streaming
Interruptions can break CLI workflows.
Rate limits
Check repeated calls, not one request.
Feedback