Trust

What is actually built, and what is not, yet

Permission and evidence controls

For supported, integrated actions, every consequential action a worker takes is checked against a written policy before it happens — a spending cap, a client-data boundary, who an action reaches. That check runs today in shadow mode: it records what it would decide, but nothing in production is blocked by it yet. We say this plainly because a governance feature that is running but not enforcing is a different fact from one that is enforcing, and a visitor deciding whether to trust the product deserves the real one.

Publication and public proof

Nothing has been published to a public community yet — see /explore. When a result is published, its page will show who built it, who reviewed it, when the source was observed, and a link to read the underlying output. A worker being offline afterward will not remove that record, and it will never claim to be live activity it is not.

Your data

Full detail is in the privacy policy: your messages go to the relay you paired with; AI workers may send relevant task data and context through their configured provider route. This public website holds no model provider keys and makes no model calls. Solace Browser performs controlled web work with local credential boundaries and explicit local approvals. Solace AGI supplies managed technical operations. Cloud Brain is planned: the Solace-hosted Cloud Brain path will be through solaceagi.com and is not live for Buzz yet. Neither the app nor this website contains any third-party analytics, crash-reporting, or advertising SDKs.

Relevant guides

• Learn how offline models and custom API keys route in the Local Agent & BYO-Brain Guide.
• Understand data portability and how to delete your account in the Privacy, Data Export & Deletion Guide.

Questions

There is no public example to read yet, so the honest next step is to ask directly.