Trust & Safety
Safety model built for real operational workflows.
DoWhiz agents run work across tools and channels with clear access controls. This page explains how isolation, permissions, and data handling work in plain language.
Last updated: February 25, 2026
Isolated execution
Tasks run in isolated execution environments designed to reduce blast radius and keep runtime boundaries explicit. Isolation controls are part of our default workflow model, not an optional add-on.
Permission model
- Agents use agent-owned identities/accounts.
- DoWhiz does not require your personal credentials for external tools.
- Agents access workspaces only when you explicitly grant permissions.
- You can revoke access through your workspace/platform controls.
Data handling basics
- Data is used to route, execute, and return requested work outputs.
- Shared memory stores per-user context so follow-up requests stay aligned.
- Retention follows operational and legal requirements.
- Deletion requests can be submitted via admin@dowhiz.com.
Auditability and transparency
Requests and outputs are tied to their trigger surface (for example: issue, message, or comment thread), which supports transparent handoffs and traceable delivery.