Offline-first, local by default
Mira runs on your machine. You should be able to use Mira without a cloud account.
Mira is built to be calm, private, and predictable. That means clear boundaries: what Mira does, what she doesn’t do, and what you control.
These principles guide how Mira is built and how she behaves.
Mira runs on your machine. You should be able to use Mira without a cloud account.
Mira only stores information when you explicitly tell her to. You stay in control.
Minimal surface area. Stability and trust outweigh feature count.
Packs use manifest + hash verification so tampering is detected and rejected.
If something goes wrong, Mira can enter recovery mode to protect stability.
Day-1 scope is intentionally limited. Mira should be predictable and honest.
Capabilities included in Day-1 scope.
These are intentionally excluded at launch.
Memory is selective, explicit, and user-controlled.
Mira does not silently store details about you. If something should be remembered, Mira asks — or you explicitly instruct her to save it.
Profile (stable preferences), Working (short-term), and Project notebooks (durable). You control what gets stored and where.
Sensitive state is encrypted locally (OS-native encryption). You can view, delete, export, or wipe memory.
Day-1 defaults are designed to be safe and easy to reason about.
Mira binds to 127.0.0.1 by default. LAN access is off by default.
Local API calls can be protected with a key (especially for Companion ↔ Server).
Signed packs use manifest + hash verification. Tampering is rejected.
If repeated boots fail, Mira enters recovery mode so you can safely diagnose.
Common questions about trust and safety.
If you have questions about privacy or security, reach out — we’ll be direct.