RayBytes/ChatMock

What it solves

ChatMock solves the problem of being locked into OpenAI's own chat interface. It lets you use ChatGPT's Codex coding models inside your preferred chat apps, IDEs, and coding tools — anything that speaks the OpenAI or Ollama API. It acts as a local bridge so you can chat with Codex from Raycast, Terax, or any compatible client.

How it works

You install ChatMock (via Homebrew, pipx, or Docker), then log in with your ChatGPT account. After you run chatmock serve, it starts a local server on your machine (default at http://127.0.0.1:8000). That server exposes an OpenAI-compatible endpoint at /v1, and also Ollama-compatible routes. You point your existing tools at that address, and they think they're talking to a standard OpenAI server — but actually ChatMock forwards requests to your ChatGPT account and returns results. It automatically syncs the list of models available to your account, so tools can discover models like gpt-5.4 or gpt-6-astra. It supports advanced features like tool calling, structured outputs, vision input, and configurable reasoning.

Who it's for

Developers and power users who already have a ChatGPT subscription and want to use Codex models outside the official ChatGPT web/terminal UI — for example, inside Raycast, Terax, or their own OpenAI-based workflows. It's also helpful for people running headless servers (using --headless flag).

Highlights

• Works with any OpenAI-compatible client (base URL http://127.0.0.1:8000/v1)
• Also exposes Ollama-compatible endpoints
• Supports tool/function calling, structured JSON outputs, vision, and web search
• Configurable reasoning effort, thinking summaries, and fast mode
• Automatically discovers the models on your ChatGPT account
• Easy install via Homebrew, pipx, or Docker