Mission

Build an AI companion capable enough to be trusted with real work — and honest enough that the trust is earned, not assumed.

Vision

A desktop where the boundary between “asking an AI for help” and “the work actually getting done” disappears — without giving up the confirmation gates and honesty that make that safe.

Values

• • Honesty over polish. Every feature and runtime doc on this site lists real limitations next to real capabilities.
• • Confirmed, not silent. Destructive and production-impacting actions are always gated.
• • Outcome-based trust. Autonomous Engineering Task billing only fires on genuine completion — never on effort alone.

Why PawOS

Because most AI products optimize for a good demo. We'd rather optimize for the version that's still trustworthy after you've used it for six months on real work.