codex-service

English | 繁體中文

A lightweight Go proxy that authenticates with OpenAI Codex via ChatGPT OAuth and exposes OpenAI-compatible API endpoints. Use your ChatGPT Plus/Pro subscription to power any tool that supports the OpenAI API.

Disclaimer: This project uses the same public OAuth endpoints and Codex API as the official OpenAI Codex CLI. However, there is no guarantee that this usage complies with OpenAI's Terms of Service. Use at your own risk. The author assumes no responsibility for any account suspension or other consequences resulting from the use of this tool.

Features

• ChatGPT OAuth authentication — Device code flow, no API key needed
• OpenAI API compatible — Works with any tool that supports the OpenAI API (Cursor, aider, Continue, etc.)
• Chat Completions → Responses API translation — Automatically converts between formats
• SSE streaming — Full streaming support with format translation
• Auto token refresh — Handles OAuth token lifecycle automatically
• Zero external dependencies — Pure Go stdlib

Quick Start

Build & Login

go build -o codex-service .
./codex-service login

Follow the instructions to authenticate with your ChatGPT account at auth.openai.com/codex/device.

Run

./codex-service

The service starts on http://localhost:8787 by default.

Docker

# Login first (on host)
go run . login

# Run with Docker Compose
docker compose up -d --build

API Endpoints

Endpoint	Method	Description
/v1/chat/completions	POST	OpenAI Chat Completions compatible endpoint
/v1/responses	POST	OpenAI Responses API passthrough
/v1/models	GET	List available Codex models
/health	GET	Health check

Available Models

• gpt-5.4
• gpt-5.4-mini
• gpt-5.3-codex
• gpt-5.2-codex
• gpt-5.2
• gpt-5.1-codex
• gpt-5.1-codex-max
• gpt-5.1-codex-mini

Usage Examples

curl

# Non-streaming
curl http://localhost:8787/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "hello"}
    ]
  }'

# Streaming
curl http://localhost:8787/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "messages": [{"role": "user", "content": "hello"}],
    "stream": true
  }'

Python (OpenAI SDK)

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8787/v1", api_key="anything")
resp = client.chat.completions.create(
    model="gpt-5.4",
    messages=[{"role": "user", "content": "hello"}],
)
print(resp.choices[0].message.content)

Tool Configuration

For any OpenAI-compatible tool (Cursor, Continue, aider, etc.):

• Base URL: http://localhost:8787/v1
• API Key: any value (or set CODEX_LOCAL_AUTH for authentication)
• Model: gpt-5.4 (or any model from the list above)

Supported Parameters

Parameters sent to the Codex API:

Chat Completions Parameter	Codex Mapping	Notes
messages[role=system]	instructions	Required, defaults to "You are a helpful assistant."
messages[role=user/assistant]	input	Conversation messages
model	model
tools	tools
tool_choice	tool_choice
response_format	text

Parameters silently ignored (not supported by Codex Responses API):

temperature, top_p, max_tokens, max_completion_tokens, stop, frequency_penalty, presence_penalty, seed, user

Automatically set by the proxy:

Parameter	Value
stream	true (Codex requires streaming)
store	false
reasoning.effort	"medium"
reasoning.summary	"auto"

Configuration

Environment Variable	Default	Description
CODEX_LISTEN_ADDR	:8787	Listen address
CODEX_DATA_DIR	~/.codex-service	Credentials storage directory
CODEX_LOCAL_AUTH	(none)	Optional bearer token for incoming requests

How It Works

Client (Cursor, aider, etc.)
  │ POST /v1/chat/completions
  ▼
codex-service (localhost:8787)
  │ 1. Extract system messages → instructions
  │ 2. Transform Chat Completions → Responses format
  │ 3. Inject OAuth token + ChatGPT-Account-Id header
  ▼
chatgpt.com/backend-api/codex/responses
  │ SSE streaming response
  ▼
codex-service
  │ Transform Responses → Chat Completions format
  ▼
Client receives OpenAI-compatible response

Disclaimer

This project is provided as-is for educational and personal use. It uses the same public OAuth endpoints as the official OpenAI Codex CLI, but is not affiliated with or endorsed by OpenAI. The author assumes no responsibility for any account suspension, data loss, or other consequences resulting from the use of this tool. Use at your own risk.

License

MIT