A drop-in endpoint for the Codex CLI
Set the base URL and API key once and the Codex CLI keeps its normal workflow. No forked client, no patched config — the gateway speaks the same Responses API shape Codex already sends.
Codex API
Codex API & Codex CLI gateway
Keep Codex running on a stable endpoint you control.
Point the OpenAI Codex CLI at one gateway and keep coding. Requests are forwarded to healthy upstream channels, logged per request, and billed from plans or wallet balance instead of a per-seat subscription.
Failover keeps long coding sessions alive
Codex sessions are long-running, so a single upstream hiccup can waste a whole task. The gateway watches channel health and reroutes to a healthy upstream automatically, which keeps interactive sessions from dying mid-edit.
Pay for what a coding session actually costs
Buy day, week or month plans, or draw from wallet balance. Every request records model, tokens and cost, so you can see what an agent run really spends instead of guessing against a flat subscription.
Does it work with the official OpenAI Codex CLI?
Yes. Configure the gateway base URL and an API key from your console; the Codex CLI works unchanged because the gateway supports the same /v1/responses interface.
What happens when an upstream channel degrades?
The gateway marks the channel unhealthy and routes new requests to another channel in the same group, so an unstable upstream does not end your session.
How is Codex usage billed?
Through subscription plans with daily or total quotas, or from wallet balance. Usage logs report tokens and cost per request.
One API, multiple choices
Claude Code API & Claude API gateway
Stable Anthropic Messages access for Claude Code and agents.
GPT API & OpenAI Responses gatewayLatest GPT, Responses and Chat Completions through one endpoint.
AI API relay, explainedWhat a relay does, how to judge one, and how ours works.
OpenAI-compatible API gatewayOne endpoint, many models, automatic failover.