Filter by agent
Across every agent in your tenant, narrow to a specific runtime, version, or experiment cohort.
Filters on the left. Span-tree drill-down in the middle. Receipt detail and integrity proof on the right. Built for the question "what did this agent do to this customer between these dates" — and for nothing else.
Across every agent in your tenant, narrow to a specific runtime, version, or experiment cohort.
Reconstruct the full span tree of a session, including sub-agent and tool calls.
The primary CISO query — every action the agent took for a specific data subject.
"Show me everything that touched PII / PCI / financial / biometric in this period."
Auto-approved, routed to human review, denied, escalated. Composable with any other filter.
Surface actions taken outside scope, after a revocation, or under a delegation that has since changed. Was the agent actually allowed to do this — as of the moment it acted?
Failed tool calls, policy denials, model timeouts. Surfaces material events automatically.
From any receipt an operator can hand out the as-of record: a single signed, independently timestamped artefact that separates it happened (tamper-evidently) from it was authorised (in scope, delegated, not revoked). It reproduces byte-for-byte offline, so the recipient can verify it without access to Agent Audit.
The dashboard is live now against a development tenant — sign in with the dev API key and explore.
Open the dashboard