Trust is earned in steps. A new coworker starts by watching. When its drafts hold up, it suggests — a person approves every send. When the record is proven, it executes inside policy, and exceptions still route to review. You move the dial per agent and per workflow, and you can halt any agent instantly, without a ticket.
Monthly budget
$763 of $1,200
Access
Zendesk · Docs · Slack
Least privilege · read + draft only
Every level writes the same audit record, so moving the dial never costs you the trail.
Budgets, access scopes, approval requirements, and autonomy thresholds live as policy code (Cedar — the same policy language AWS uses), versioned next to the agent it governs. The platform enforces it on every run; nobody has to remember it.
Any operations lead can pause a single agent or halt it entirely from the registry. In-flight work stops at the next checkpoint, pending approvals stay in the inbox, and the record shows who pulled the switch and when.
The open-source Agent Workforce Platform. AI coworkers you can trust with real work — running in your own AWS account, on a foundation your company owns.
MPL-2.0 · Self-hosted · Runs on AWS