No telemetry.
Umbratic does not collect usage analytics or automatically upload telemetry. Task history and diagnostic logs stay on your device. You can choose to export logs when troubleshooting.
Privacy by design.
Choose where your models run. Keep control of what leaves your workspace.
Local means local inference.
Local mode runs the model on your computer through the bundled Veil gateway. There is no Umbratic account to create. Choose and download a model in the app; its inference runtime is included.
Your server. Your policy.
Company server mode sends model inputs to your configured Veil server. Its administrator manages access, model infrastructure, and server logs. Server privacy and retention depend on that deployment.
Cloud is your choice.
Connect ChatGPT through the official Codex runtime or Claude through Claude Code. Requests and selected workspace context go to that provider under your account and its policies. Those conversations are not local inference.
Connections have a purpose.
Model downloads, GitHub release checks, enrollment, and explicitly used web search require network access. Workspace commands can also access the network according to your permissions.
The source is open.
This website loads its assets from this site, with no analytics scripts or embedded contact service. You can inspect the app’s implementation and report problems on GitHub.
Read the source