macOS
Apple Silicon + Intel
Gatekeeper may warn about an unsigned app until Developer ID signing is configured — right-click → Open to launch. Updates are manual for now; new versions land on this page.
Free desktop app
TracingInsights for Desktop
Run the full F1 statistics platform in its own window with a built-in AI inference harness for ChatGPT Codex, Grok (𝕏 Premium+ / SuperGrok), and your own local or hosted OpenAI-compatible models — your credentials stay on your machine. Available for macOS, Windows, and Linux.
Download by platform
Windows
Installer
SmartScreen may show an "unknown publisher" warning — choose More info → Run anyway. The installer offers a full setup flow with a selectable installation directory.
Linux
AppImage · .deb on the releases page
The AppImage is the quickest path (make it executable and run). A .deb package is also published on the releases page for Debian/Ubuntu users.
All installers live on the public GitHub releases page — .deb, .zip, and release metadata included.
Why run the desktop app?
Same TracingInsights you know — plus a built-in inference harness that keeps your Codex and Grok calls (and your own models) on your own machine.
-
Codex & Grok through your own machine
The desktop app runs the inference protocol layer in its own main process, so ChatGPT Codex and Grok (𝕏 Premium+ / SuperGrok) calls go through your computer instead of the TracingInsights server.
-
Your tokens never leave your device
ChatGPT and 𝕏 credentials are stored in the app’s secure system keyring (safeStorage). The desktop app owns them end-to-end; the TracingInsights server never sees them.
-
Bring your own models
Add a local model (Ollama, LM Studio, vLLM, llama.cpp) or a hosted OpenAI-compatible endpoint (Fireworks, Together, DeepInfra, …) and use it for chat — free, no subscription gate.
-
No browser CORS roadblocks
Browsers cannot call chatgpt.com or cli-chat-proxy.grok.com directly. The desktop app’s main process makes those calls natively, with no CORS or loopback bridge.
-
Automatic updates
Shipped apps check the public releases repo and update in place. Windows and Linux auto-update today; macOS joins as soon as Developer ID signing is configured.
-
One installer for every Mac
The macOS build is a universal binary — a single download runs on both Apple Silicon and Intel.
Getting started
Install the app, connect ChatGPT Codex or Grok with a device code, and optionally bring your own model endpoint.
-
Install the app
Pick your platform above and open the installer. The app loads TracingInsights in its own window.
-
Sign in to your provider
Connect ChatGPT Codex or Grok (𝕏 Premium+ / SuperGrok) in Settings → AI Provider, as you would in the browser.
-
Bring a custom provider (optional)
In Settings → AI Provider → Custom providers, add a local model (Ollama, LM Studio, vLLM, llama.cpp) or a hosted OpenAI-compatible endpoint. Codex and Grok connect with a device code — the server never sees your tokens.