ALLOW
Routine action satisfies policy and proceeds.
Product
Independent authorization before AI-agent actions execute
BobSentry is not an orchestration framework, observability platform, chatbot guardrail, workflow automation product, or generic AI governance dashboard. It is the authorization boundary immediately before an action reaches a real system.
- 01AI agent proposes an operational action
- 02BobSentry runtime evaluation
- 03ALLOW / REQUIRE_APPROVAL / BLOCK
- 04Only authorized actions continue
- 05Signed decision evidence recorded
BobSentry decides before the external system is changed.
Runtime decision contract
Every consequential evaluation returns one of three decisions.
REQUIRE_APPROVAL
A consequential or ambiguous action is held for operator authorization.
BLOCK
A prohibited action is stopped before execution.
Approvals bound to the exact proposed action
When BobSentry returns REQUIRE_APPROVAL, an operator authorizes or rejects the specific action—target, metadata, policy version, and expiration included. A standing permission is not enough.
- 01Agent proposes an operational action
- 02BobSentry evaluates before execution
- 03Returns ALLOW / REQUIRE_APPROVAL / BLOCK
- 04Approval is requested when required
- 05Only authorized actions continue
- 06Signed evidence is recorded
BobSentry evaluates before execution. Only authorized actions continue, with signed decision evidence recorded.
Observe
Evaluate actions and collect evidence without blocking execution. Use this mode to validate policies and coverage before gating.
Enforce
Activate approval gating and blocking before consequential actions execute. Roll out when your interception path and policies are ready.
Observe, then Enforce
Start in Observe to validate policy coverage and interception quality. Move to Enforce when you are ready for approval gating and blocking.