LOCAL AI · HUMAN-GOVERNED
Your AI assistant,
running on your machine
Yana AI Desktop brings the same governed runtime to your machine: bring your own API key, or run fully offline with Ollama. No Yana AI server ever sits between you and your data, and every safety hook still runs locally, under your control. 2,025 skills, 100 agents, deterministic policy gates.
Free · Open source (Apache 2.0) · All releases →
- Works offlineOllama · local models
- Multi-layer securityDeterministic policy gates
- Fully privateNo telemetry
- Bring your own API keyDirect to provider
Everything you need, running on your machine
No cloud, no tracking. Your data stays with you.
-
Multi-layer security
Runtime hooks enforce policy before and after sensitive AI actions.
L0–L5Policy layers -
Bring your own API key
Use your own Anthropic, OpenAI, Gemini, or Ollama key. No middleman — pay the provider directly.
-
2025 built-in skills
An extensive skill library — code review, RAG, multi-agent orchestration, cybersecurity, and business ops.
-
Works offline
Connect Ollama to run local models (Llama, Qwen, Mistral). No internet required, no API costs.
-
Smart dashboard
Track usage, agents, memory, and tasks in one unified interface — fast and beautiful.
-
Fully private
No analytics, no telemetry, no conversation sharing. Open source — verify it yourself.
Common questions
Still deciding? Here's what people usually ask first.
Do I have to pay for Yana AI Desktop?
The app itself is free and open source (Apache 2.0). You bring your own API key from Anthropic, OpenAI, Gemini, or a local model — you only pay the provider directly for what you actually use, at their normal rate. No subscription, no markup.
Is my data actually private?
The app runs entirely on your machine — no Yana AI server sits between you and your chosen provider. There's no analytics, no telemetry, and no conversation logging on our end. Since it's open source, you can read the code yourself instead of taking our word for it.
I don't have an API key — can I still use it?
Yes. Connect Ollama and run an open-weight model (Llama, Qwen, Mistral) fully offline, with no key and no cost. Cloud providers are optional, not required.
What does "multi-layer security" actually block?
Runtime hooks check sensitive actions before and after execution — including prompt injection attempts, destructive shell commands, and secrets accidentally leaking into output. It's enforced code, not a policy document.
I see both an old and a new interface — which one am I in, and how do I switch?
The new workspace shell is the default and where active development happens. If you land in the legacy interface, open its Settings page → "Interface" card → the button there switches you back to the new workspace. From the new workspace, use the "•••" menu in the top-right header → "Legacy UI" to go the other way. Either choice is remembered the next time you launch the app, so you won't need to switch again until you choose to.
Available on every platform
v1.4.8 for every platform — download and run instantly, no extra setup needed.
-
macOS
Apple Silicon (arm64) · Intel (x64)
Download .dmg (arm64) → Download .dmg (Intel) →- Open the .dmg, drag "Yana AI" into Applications.
- Open the app from Applications (not from inside the .dmg).
- Ad-hoc signed, not notarized by Apple yet, so macOS will warn "unidentified developer" — right-click (Control-click) the app → Open → Open again in the dialog.
- If macOS already refused to open it: go to System Settings → Privacy & Security → scroll to Security → click "Open Anyway" next to Yana AI, then open it again.
- Full walkthrough (and why it's not notarized yet): docs/MACOS_INSTALL.md
-
Windows
x64 · Installer (.exe)
Download .exe →- Run the downloaded .exe.
- Not code-signed yet, so Windows SmartScreen will warn — click "More info" → "Run anyway".
- Follow the installer, then launch Yana AI from the Start menu.
-
Linux
x64 · arm64 · AppImage
Download AppImage (x64) → Download AppImage (arm64) →- Make it executable:
chmod +x Yana-AI-1.4.8-x86_64.AppImage - Run it:
./Yana-AI-1.4.8-x86_64.AppImage - If it complains about missing FUSE (common on newer Ubuntu/Fedora), install
libfuse2, or run with--appimage-extract-and-runinstead.
- Make it executable:
Get started today
Download, enter your API key, and you're ready — takes less than 2 minutes.