Agent Inventory
Owner, mission, autonomy, SOR/API dependencies, policies, inputs, outputs, certification posture, and runtime evidence.
Open inventorybp Agent Governance & Replay Control Tower
Governable registry, engineering trace, evaluation, policy gateway, knowledge stewardship, and replay.
Governable agent estate, not opaque automation
This surface proves the runtime claims in validation. It shows who owns each agent, what stage it is in, how it is certified, how policy gates intervene, how evidence and replay are attached, and how people govern the estate without leaving context.
BP technology audit coverage
This is the enterprise-agent-runtime answer to the validation questions: what agents exist, what evidence proves them, how decisions replay, what policies govern actions, who owns runtime risk, and where operations teams monitor it.
| Link | Artifact | Controls mission parity proof |
|---|---|---|
| Signal | TICK-GOV-77102 | Agent permission, evidence completeness, replay coverage, or certification state changed. |
| Context | CTX-GOV-77102 | Agent identity, owner, scope, skill, SOR access, policy binding, and runtime status resolved. |
| Policy | FIN-AI-GOV-001 v2.0 | Owner, permission, human boundary, replay, evidence contract, and certification controls enforced. |
| Evidence | EP-GOV-77102 | Registry row, execution record, policy binding, access contract, skill dependency, and eval result attached. |
| Decision | DEC-GOV-77102 | Certify, restrict, retire, or remediate agent based on governance posture. |
| Action | ACT-GOV-77102 | Create certification task and restrict unsafe action permissions until owner approval. |
| Outcome | 100% high-risk agents owner-bound | Governance risk reduced before production action; human accountability remains explicit. |
| Replay | RPL-GOV-77102 · 99.3% | Agent identity, evidence, policy, action boundary, certification decision, and learning signal replayable. |
| Learning | LRN-GOV-77102 | Certification gap and policy-denial patterns are retained for future agent readiness scoring. |
Evidence Intelligence
Evidence vault and packs expose source proof, lineage, integrity, and transaction-level records instead of summary-only claims.
Open evidence vaultReplay Explorer
Decision Replay Studio reconstructs signal, context, evidence, policy, agent output, approval, action, and outcome where records exist.
Open replayRuntime Control Plane
Runtime operations surface shows runtime health, governance posture, replay readiness, SOR fabric, and production swap readiness.
Open runtime opsPolicy Runtime
Policy registry and trust fabric expose deterministic controls, policy coverage, drift, approval boundaries, and write-path gates.
Open policiesAgent Certification
Certification state is visible through agent specs, evaluation gates, assurance operations, promotion stage, and rollback metadata.
Open assuranceOwnership
Business owner, technical owner, first-line, second-line, and third-line governance appear in inventory and drawer details.
Open ownershipHuman Accountability
Recommendations remain tied to policy gates, approval chains, escalation rationale, action boundaries, and human supervision.
Open autonomy governanceData Governance
Source systems, access path, steward, UDP contract, semantic model, freshness, and retrieval audit are shown for agent knowledge sources.
Open context graphObservability
Business observability covers policy gates, evidence coverage, replay, evaluation, drift, autonomy posture, and runtime health.
Open runtime command