Use OpenAI Codex with your own local model

OpenAI Codex talks to Tokios through the Responses API. Add Tokios as a model provider in ~/.codex/config.toml, point it at https://api.tokios.com/v1, and Codex runs against your own local model — on Ollama, LM Studio, or any OpenAI-compatible backend — from anywhere, down your private tunnel.

Prerequisites

A running Tokios connector paired with your local model — see Connector install
A registered model deployment name — see Register a model
A Tokios API key — see API Keys
OpenAI Codex installed

Configure Codex

Set your Tokios key in the environment, then add a Tokios provider and profile to ~/.codex/config.toml:

export OPENAI_API_KEY=sk-tok-YOUR_KEY

$env:OPENAI_API_KEY = "sk-tok-YOUR_KEY"

~/.codex/config.toml

[model_providers.tokios]
name = "Tokios"
base_url = "https://api.tokios.com/v1"
env_key = "OPENAI_API_KEY"
wire_api = "responses"

[profiles.tokios]
model = "gemma-tunnel"
model_provider = "tokios"
disable_response_storage = true

wire_api = "responses" and disable_response_storage = true are both required. Tokios serves the Responses surface statelessly, so Codex must not rely on server-side response storage.

Then run Codex with the Tokios profile:

codex --profile tokios

Replace gemma-tunnel with the public model name you registered in the console.

Using the OpenAI Python SDK

To call your local model programmatically with the OpenAI SDK, point the client at the Tokios base URL:

openai_tokios.py

from openai import OpenAI

client = OpenAI(
    base_url="https://api.tokios.com/v1",
    api_key="sk-tok-YOUR_KEY",
)

response = client.chat.completions.create(
    model="gemma-tunnel",
    messages=[{"role": "user", "content": "Explain async/await in Python"}],
)
print(response.choices[0].message.content)

The OpenAI SDK’s base URL must end with /v1. The Tokios endpoint is https://api.tokios.com/v1.

Troubleshooting

401 Unauthorized

Your sk-tok-… key is missing or wrong. Confirm OPENAI_API_KEY is set in the environment and that env_key = "OPENAI_API_KEY" in config.toml matches. Check the key on the Keys tab — it was shown only once at creation.

404 Not Found

The model in your config.toml profile does not match any deployment name. Check the Models tab for the exact name you registered — it is case-sensitive.

503 Service Unavailable

The connector for that deployment is offline or at its concurrency cap. Check the Connectors tab — the connector needs to show as Online, and the machine running it needs to still be up.

Codex ignores the Tokios profile

Confirm you are passing --profile tokios on the command line. Without the flag, Codex uses its default profile, which points at OpenAI’s hosted API. You can also set default_profile = "tokios" under [general] in config.toml to make it the default.

Next steps

Compatible clients

See base URL and key setup for other OpenAI- and Anthropic-compatible clients.

API keys

Create, scope, and rotate sk-tok-… keys for teammates and agents.

Getting started with local LLMs

Hardware & fit

Serve & connect your runtime

Connect a client or app

Compare

Use OpenAI Codex with your own local model

Prerequisites

Configure Codex

Using the OpenAI Python SDK

Troubleshooting

Next steps

Compatible clients

API keys

​Prerequisites

​Configure Codex

​Using the OpenAI Python SDK

​Troubleshooting

​Next steps

Compatible clients

API keys

Prerequisites

Configure Codex

Using the OpenAI Python SDK

Troubleshooting

Next steps