Platform / Governance
What an agent is allowed to do, made explicit.
Governance sits between a proposed action and its execution. Low-risk actions matching policy proceed; everything else routes to a named reviewer.
How a decision routes
Policy first, then either straight through or a human decision.
Scroll to see the full diagram →
Action Requested
An agent proposes an action — a tool call, an update, a message.
Key concepts
The building blocks of governance.
Policy
Rules that define which actions are permitted, restricted, or require review.
Approval thresholds
Risk levels that determine whether an action proceeds automatically or waits for a person.
Human checkpoints
Named reviewers with the authority to approve, edit, or deny a specific class of action.
Boundaries
The explicit limits of what an agent can touch — systems, records, and actions it cannot reach.
Limitations
Governance is configured, not assumed.
Policy and approval thresholds are defined for your workflow during design — the platform does not ship with a generic risk model that applies to every business.
A human checkpoint is only as effective as the reviewer behind it. The platform surfaces the decision; the operating model around it is part of the engagement.
Define the policy for one workflow.
Bring an action you'd want gated — a payment, a record change, a message sent externally — and we'll map the checkpoint.