Private AI for Windows
Private local AI for Windows.
Use AI models on your own computer, keep your conversations local, and review protected actions before they run.
From download to first chat
From download to your first private chat.
Review the unsigned build, choose a model, then start a local chat.
- 01Review and install the Windows build
Check the unsigned publisher status, exact version, and SHA-256. Install only if the hash matches and your device policy permits it.
- 02Choose a local model
Model browsing and downloads use the network only when you start them. CPU operation is verified; GPU acceleration is not yet verified.
- 03Chat on your computer
Prompts, replies, and local memory stay on this PC. Protected requests pause for a local decision.
Privacy without ambiguity
What stays local—and what uses the network.
Core local chat needs no website account or optional plugin after the app and model files are present.
Private work stays with the app
- Prompts, replies, and local memory
- Downloaded model files
- Protected-request decisions
Network use stays visible
- Website and release information you view
- Model catalog checks and downloads you start
- Remote requests after RemoteEndpoints setup and local approval
- Public-safe memory summaries after MemoryEndpoints setup and local enablement
MemoryEndpoints hosted save and search are not yet verified end to end for this build. Keep important memory local.
Current early access
Review version 0.2.242.0 before you install.
The Windows x64 build is unsigned, so Windows can show Unknown publisher. Compare the SHA-256 and follow your device policy. This is first-party product documentation, not an independent audit or certification.