Onyx wraps Claude Code, Codex, Gemini CLI and Kimi Code in a single Rust-native GUI — with Git, terminal, diffs, provider chats and system-wide voice built in.
macOS · Windows · Linux — bring your own provider logins
Onyx listens system-wide. Dictate straight into your editor, terminal or browser — or summon a full voice agent that speaks back. No window focus required.
Stop juggling terminals and browser tabs. Onyx normalizes every agent's events into one consistent, fast UI.
Run Claude Code, Codex, Gemini CLI and Kimi Code side by side. Onyx reuses each tool's own login and runtime — nothing to reconfigure, nothing replaced.
Working-tree changes, branch switching, commit & push, file tree, integrated terminal and an in-app browser — no context switching.
Long sessions stay navigable. Compact markers per prompt with hover previews and click-to-jump, plus steering, approvals and Build/Plan controls.
ChatGPT, Claude, Gemini and Grok open in internal webviews that keep your own sign-ins — plus a native OpenRouter route with model discovery and web search.
System-wide dictation and a hands-free voice agent with floating HUDs — Onyx hears you in any app, not just its own window.
Sessions, history and preferences live on your machine. API keys stay in the OS credential manager. Optional cloud sync when you want it.
Built on Tauri 2 with a Leptos + WebAssembly frontend — a tiny footprint, instant startup and signed auto-updates. One language, from shell to UI.
MIT-licensed. Bring your own provider subscriptions or API keys — Onyx never sits between you and your agents.
Independent project — not affiliated with Anthropic, OpenAI, Google, Moonshot or xAI.