Codex (OpenAI CLI)

Run new local Codex tasks through OneProvider's streamed Responses API.

Codex coding agent CLI and desktop app

What works

Supports streamed responses, function calls, local tools such as apply_patch, and usage accounting. This changes new local CLI, desktop, and IDE tasks; it does not change Codex cloud chats.

Client limitation

Restart Codex and create a new local task after changing the provider. Existing tasks keep the model and provider they started with.

Setup

1. Install or verify Codex

   If codex --version already prints a version, continue. Otherwise install the current CLI.

   npm install --global @openai/codex
   codex --version

2. Make a dedicated key available

   Create a dedicated OneProvider key and export it in the same terminal session that starts Codex.

   export ONEPROVIDER_API_KEY="sk-your-key-here"

3. Add the custom provider

   Open ~/.codex/config.toml (desktop: Settings → Configuration → Open config.toml) and add the OneProvider provider.

   model_provider = "oneprovider"
   model = "gpt-6.1-sol"
   [model_providers.oneprovider]
   name = "OneProvider"
   base_url = "https://api.oneprovider.dev/v1"
   env_key = "ONEPROVIDER_API_KEY"
   wire_api = "responses"

4. Restart and test a new local task

   CLI users should run a read-only probe. Desktop users must fully quit, reopen, and start a new local task.

   codex

Verify the connection

Confirm the request in your dashboard

The completed request should show the selected model, /v1/responses, token usage, and a settled charge.

Connection details

Protocol

responses

Endpoint

https://api.oneprovider.dev/v1

Key security

Use a dedicated scoped key. On macOS, a desktop app opened from the Dock may not inherit terminal environment variables; use Codex command-backed Keychain authentication instead of placing the secret in config.toml.

Which model to choose

Set model in the same config.toml, below model_provider. gpt-6.1-sol is the frontier tier for hard agentic tasks, gpt-5.6-terra is the balanced default for everyday work, and gpt-5.6-luna is the low-cost tier for high-volume and quick tasks. Codex talks the Responses API, so pick GPT ids from the models page ; rates are on the pricing page .

Common errors and fixes

401 or “missing credentials”

ONEPROVIDER_API_KEY is not visible to Codex. Export it in the terminal that starts Codex. A desktop app opened from the Dock does not inherit it on macOS: use command-backed Keychain authentication.

404

base_url must be https://api.oneprovider.dev/v1 with /v1, and wire_api must be "responses".

Codex still uses the old model or provider

Existing tasks keep the provider they started with. Restart Codex and create a new local task.

“You've hit your usage limit” or Codex asks me to sign in to ChatGPT

That is the limit of a ChatGPT subscription: the request went to ChatGPT, so the OneProvider provider is not active. Put model_provider = "oneprovider" at the very top of config.toml, above any [table] (no smart quotes), quit Codex fully and start a new local task. Then sign out of ChatGPT with codex logout (desktop app: in its settings) and start codex in the terminal where ONEPROVIDER_API_KEY is exported: with the provider active it works on the key alone. Quick test: codex -c model_provider=oneprovider. wire_api accepts only "responses".

429 “Insufficient balance”

Balance or key limit is used up. Top up; retries will not help.

FAQ

Does this change Codex cloud chats?

No, only new local CLI, desktop and IDE tasks.

Do I need an OpenAI account?

No, the OneProvider key replaces it for local tasks.

Can I use Claude models in Codex?

Codex uses the Responses API with GPT ids. For Claude use Claude Code .

Where do I check cost?

In the dashboard Usage page: each request shows the model, /v1/responses and a settled charge.

Quick answers

Which base URL do I use?

Claude Code and Anthropic SDKs use https://api.oneprovider.dev. OpenAI SDKs and OpenAI-compatible clients use https://api.oneprovider.dev/v1. One key works behind both; the model list is on the models page ; rates are on the pricing page .

How do I authenticate?

Send the raw sk-… key in x-api-key or Authorization: Bearer: one of them, never both. Keep the key in an environment variable, not in a file that gets committed. More on authentication .

Where do I get a key?

Open your cabinet in the Telegram bot @oneprovider_robot or at app.oneprovider.dev, top up the balance and press “Issue a key”. Each key can carry its own spend limit. Cabinet & API keys .

What do 401, 404 and 429 mean?

401: the key is missing, mistyped or sent in the wrong header; paste the whole key without quotes or spaces. 404: usually a wrong base URL, for example a missing or doubled /v1. 429 with “Insufficient balance”: the cabinet balance or the key’s spend limit is used up, so top up instead of retrying; any other 429 is a rate limit, so wait for Retry-After. Details: error codes and troubleshooting .

Why do streams cut off or reset?

VPN and VLESS proxies often buffer SSE or break HTTP/2 negotiation. OneProvider has no regional restrictions and does not buffer streams, so turn the proxy off and retry.

Where do I ask if this is not enough?

Message @Oneproviderhelpbot on Telegram or open “Support” in the web cabinet, with the HTTP status code and the exact error message. How support works .