CodexAPI

OpenAI compatible API powered by your ChatGPT plan.

A local OpenAI-compatible API proxy that routes requests through your ChatGPT account. Drop it in front of any OpenAI-compatible client and use GPT-5 and Codex models without a direct API key.

Your client  ──►  CodexAPI (:8000)  ──►  chatgpt.com
(OpenAI format)   (local proxy)         (your account)

---

Table of contents

• Quick start
• Setup
  • Option A — CLI (local)
  • Option B — Docker
  • Option C — Client library only
• Use
• Supported models
• Endpoints
• Configuration
• CLI reference

---

Quick start

# 1. Install
git clone https://github.com/kkristof200/codex-api && cd codex-api
pip install uv && uv sync

# 2. Login
codexapi login

# 3. Serve
codexapi serve

Then point any OpenAI-compatible client at http://localhost:8000.

---

Setup

Option A — CLI (local)

Note

Requires Python 3.13+ and a ChatGPT account (Plus, Pro, Team, or Enterprise).

1. Install

git clone https://github.com/kkristof200/codex-api
cd codex-api
pip install uv
uv sync

Or install only the server package with pip (no uv required):

pip install ./packages/codexapi-server

2. Login

codexapi login

Tip

On a headless machine, use --no-browser. CodexAPI will print an authorization URL — open it in any browser, complete login, then paste the redirect URL back into the terminal.

codexapi login --no-browser

3. Serve

codexapi serve

Available options

Flag	Env var	Default	Description
--host	—	127.0.0.1	Bind address
--port	PORT	8000	Port
--reasoning-summary	CHATGPT_LOCAL_REASONING_SUMMARY	auto	auto concise detailed none
--enable-web-search	CHATGPT_LOCAL_ENABLE_WEB_SEARCH	false	Enable web search by default

---

Option B — Docker

Note

Requires Docker and Docker Compose.

1. Configure

cp .env.example .env

Environment variables

Variable	Description
CHATGPT_LOCAL_HOME	Set to /data to match the Docker volume (default in example)
CHATGPT_LOCAL_REASONING_SUMMARY	auto / concise / detailed / none
CHATGPT_LOCAL_ENABLE_WEB_SEARCH	true / false
PORT	Host port to expose (default: 8000)

2. Login

docker compose --profile login run --rm --build --service-ports login

Your browser will open (or a URL will be printed). Complete login — the container exits automatically and tokens are saved to a persistent Docker volume.

Tip

If the browser redirect page fails to load, copy the URL from your browser's address bar and paste it into the terminal when prompted. This is normal when the container's port mapping isn't fully established.

3. Serve

docker compose up --build -d serve

The server is available at http://localhost:8000.

Re-authenticate when the token expires:

docker compose --profile login run --rm --build --service-ports login

---

Option C — Client library only

Install just the core library to make requests programmatically without the server:

pip install ./packages/codexapi-client

Or directly from GitHub:

pip install "codexapi-client @ git+https://github.com/kkristof200/codex-api.git#subdirectory=packages/codexapi-client"

from codexapi_client import CodexAPI, OpenAIChatCompletionAdaptorHTTP

codex = CodexAPI()
result = codex.request_http(
    model_name="gpt-5.4",
    reasoning_effort="medium",
    chatgpt_acc_id="your-account-id",
    auth_token="your-token",
    messages=[{
        "type": "message",
        "role": "user",
        "content": [{"type": "input_text", "text": "Hello!"}]
    }],
    adaptor=OpenAIChatCompletionAdaptorHTTP(),
)

---

Use

Point any OpenAI-compatible client at http://localhost:8000:

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

---

Supported models

Model	Reasoning efforts
gpt-5	low medium high
gpt-5.1	none low medium high
gpt-5.2	none minimal low medium high xhigh
gpt-5.4	none minimal low medium high xhigh
gpt-5-codex	low medium high
gpt-5.1-codex-mini	low medium high
gpt-5.1-codex	low medium high
gpt-5.1-codex-max	low medium high xhigh
gpt-5.2-codex	low medium high xhigh
gpt-5.3-codex	none minimal low medium high xhigh

Model name syntax

Reasoning effort and per-request options can be embedded directly in the model name:

gpt-5.4                                          # default effort (medium)
gpt-5.4-high                                     # explicit effort
gpt-5.4-high-[web_search:true]                   # with web search
gpt-5.4-high-[reasoning_summary:detailed]        # with reasoning summary
gpt-5.4-high-[web_search:true,reasoning_summary:concise]

---

Endpoints

Method	Path	Description
POST	/v1/chat/completions	OpenAI-compatible chat completions
POST	/v1/completions	OpenAI-compatible text completions
GET	/v1/models	List all available models

Both streaming ("stream": true) and non-streaming responses are supported.

---

Configuration

Variable	Description
CHATGPT_LOCAL_HOME	Directory where auth tokens are stored (default: ~/.chatgpt-local)
CHATGPT_LOCAL_REASONING_SUMMARY	Default reasoning summary verbosity (auto / concise / detailed / none)
CHATGPT_LOCAL_ENABLE_WEB_SEARCH	Enable web search by default (true / false)
PORT	Server port

---

CLI reference

codexapi login     Authenticate with ChatGPT and store tokens
codexapi serve     Start the OpenAI-compatible server
codexapi info      Print the currently stored account and token info

---

Credits

This project was inspired by ChatMock