A fast, local-first AI terminal for macOS, Linux, and Windows.
AZMX AI is a native AI-first terminal that works out of the box with free local models — no API key, no signup, no cloud roundtrip. It pairs a real PTY backend with a modern UI: multi-tab terminals, an integrated code editor, a file explorer, and a first-class AI side-panel.
Bring your own key from any major provider, or run free models locally via Ollama. The installer is ~10 MB. API keys live in your OS keychain. Telemetry is off by default.
The application source is proprietary and not hosted here. This repository exists only to publish release artifacts. Installers and the auto-updater manifest are attached to each tagged release.
- Free local AI without an API key. One-click setup of Ollama + curated coding models (Qwen2.5-Coder, Granite Code — all Apache 2.0). Your prompts never leave the machine.
- Or BYOK. OpenAI, Anthropic, Google, Groq, xAI, Cerebras, DeepSeek, NVIDIA NIM (hosted at build.nvidia.com or self-hosted), Azure OpenAI (any resource + deployment), and any other OpenAI-compatible endpoint (Vertex AI, Bedrock-via-LiteLLM, …) — any one is enough to start. Keys live in the OS keychain, never in disk files or
localStorage. - GPU-aware agent. When you're on an NVIDIA machine the agent can read live GPU state via
nvidia-smi— debug OOM, pick a batch size, fit a model to your VRAM. - Native, lightweight. ~10 MB installer. No Electron. No telemetry by default.
- Real terminal, real editor, real explorer. xterm.js+WebGL terminal, CodeMirror 6 editor with vim mode + inline AI autocomplete, file tree with git status badges.
- MCP-native. Bundled HTTP/SSE transport, curated catalog (GitHub, Postgres, Slack, Filesystem, …), keychain-backed secrets per server. Auto-imports
.mcp.jsonand~/.claude.json. - Claude Code interop. Reuses your
AZMX.md/CLAUDE.md/AGENTS.md,.claude/agents/*.md,.claude/commands/*.md, and.claude/settings.jsonhooks. Drop AZMX into a project where Claude Code already lives — it adopts the config.
Pick the installer for your platform from the latest release:
| Platform | File |
|---|---|
| macOS (Apple Silicon) | AZMX.AI_<version>_aarch64.dmg |
| macOS (Intel) | AZMX.AI_<version>_x64.dmg |
| Linux | AZMX.AI_<version>_amd64.AppImage, .deb, or .rpm |
| Windows | AZMX.AI_<version>_x64_en-US.msi or *_x64-setup.exe |
Need help installing? See SETUP.md for per-platform walkthroughs.
- Open AZMX AI.
- The first-run tour appears. On step "Free local AI", click Set up local AI.
- AZMX walks you through installing Ollama (~150 MB).
- Pick Qwen2.5-Coder 7B (default — ~4.7 GB, Apache 2.0). The pull streams in-app.
- When the bar turns green, the AI panel is live. Ask anything.
Prefer BYOK? Skip step 2's button and use the BYOK step instead. Both paths land at a working composer.
| Document | Purpose |
|---|---|
| SETUP.md | Install per platform, troubleshooting, first-run setup |
| MANUAL.md | Full end-to-end feature reference — every panel, every shortcut, every workflow |
| FAQ.md | Common questions: privacy, licensing, performance, model recommendations, integrations |
| CHANGELOG | What changed in each release |
AZMX checks for updates on launch. New versions land via the in-app updater — no need to revisit this page once you're installed. Signed via a public minisign key bundled with the app.
AZMX AI is free to download but proprietary. By installing the application you agree to the End User License Agreement that ships with it. The release artifacts are signed; the source repository is private.
- Manual / FAQ: see the links above.
- Bugs: open an issue against this repository — include OS, AZMX version, and steps to reproduce.