ChatMock

Allows Codex to work in your favourite chat apps and coding tools.

Install

Homebrew

brew tap RayBytes/chatmock
brew install chatmock

pipx / pip

pipx install chatmock

Docker

See DOCKER.md

Getting Started

# 1. Sign in with your ChatGPT account
# If you are running this on a headless server, append --headless
chatmock login

# 2. Start the server
chatmock serve

The server runs at http://127.0.0.1:8000 by default. Use http://127.0.0.1:8000/v1 as your base URL for OpenAI-compatible apps.

Example usage

Raycast Integration

1. Configure the Host URL
   Open your Raycast Extensions preferences, navigate to the Ollama settings section, and input the host URL (default is 127.0.0.1:8000).
   

2. Sync Your Models
   Click the Sync Models button, which will register all available models.

3. Start Chatting
   Open the Raycast AI Chat interface. You will now see model slugs which you can chat with.

Terax (Agentic Terminal) Integration

1. Configure the provider settings
   Open your Terax settings, and switch to the Models tab, add a new provider (OpenAI Compatible), and input the host URL (default is http://127.0.0.1:8000/v1), along with the model IDs you wish to use (API key may be anything).
   

2. Favourite, and start using it!
   Go back to your main chat window, select the model by going to the OpenAI Compatible icon, and clicking the model there (you may favourite it here to quickly select it the next time if you switch between models)

Supported Models

ChatMock automatically discovers the models available to the signed-in ChatGPT account. It checks the latest stable Codex release on GitHub when refreshing the catalog and uses that version for the account's model request. An optional CHATGPT_LOCAL_MODELS_CLIENT_VERSION override accepts a stable release version only. If GitHub is unavailable, ChatMock uses the last successful version or its bundled verified fallback. The model list and capabilities come from the account catalog; /v1/models and /api/tags return 503 until one is available. With --no-model-sync, they list a built-in static set instead, which may include models your account can't use. The catalog refresh interval defaults to one hour.

Features

• Tool / function calling
• Structured outputs via JSON schema (response_format, or format on the Ollama routes)
• Vision / image input
• Thinking summaries (via think tags)
• Configurable thinking effort
• Fast mode for supported models
• Web search tool
• OpenAI-compatible /v1/responses (HTTP + WebSocket)
• Ollama-compatible endpoints
• Reasoning effort exposed as separate models (optional)

Configuration

All flags go after chatmock serve. These can also be set as environment variables.

Flag	Env var	Options	Default	Description
--reasoning-effort	CHATGPT_LOCAL_REASONING_EFFORT	none, minimal, low, medium, high, xhigh, max, ultra	medium	How hard the model thinks
--reasoning-summary	CHATGPT_LOCAL_REASONING_SUMMARY	auto, concise, detailed, none	auto	Thinking summary verbosity
--reasoning-compat	CHATGPT_LOCAL_REASONING_COMPAT	legacy, o3, think-tags	think-tags	How reasoning is returned to the client
--fast-mode	CHATGPT_LOCAL_FAST_MODE	true/false	false	Priority processing for supported models
--enable-web-search	CHATGPT_LOCAL_ENABLE_WEB_SEARCH	true/false	false	Allow the model to search the web
--expose-reasoning-models	CHATGPT_LOCAL_EXPOSE_REASONING_MODELS	true/false	false	List each reasoning level as its own model
--model-sync	CHATGPT_LOCAL_MODEL_SYNC	true/false	true	Discover account models automatically
--model-refresh-interval	CHATGPT_LOCAL_MODEL_REFRESH_INTERVAL	seconds	3600	Refresh interval for model discovery

Web search in a request

{
  "model": "gpt-5.4",
  "messages": [{"role": "user", "content": "latest news on ..."}],
  "responses_tools": [{"type": "web_search"}],
  "responses_tool_choice": "auto"
}

Fast mode in a request

{
  "model": "gpt-5.4",
  "input": "summarize this",
  "fast_mode": true
}

Note

This project is not affiliated with OpenAI.

Star History