Every agent action, hash-chained and exportable.
What you can show a regulator without our help. Every input, every output, every tool call, every approval — signed, timestamped, and linked to the policy it was bound by.
Design-partner roadmap. The full agent-level trail described on this page — per-tool-call capture, source signing, one-click export, and offline replay — is in design-partner pilots. Available today: a hash-chained, tamper-evident audit of every privileged admin action, exportable from the admin console and re-verified on a schedule.
The trace covers everything that touched the agent.
Inputs & system msgs
Full text, model, parameters, version of the prompt template. PII redacted in clear text.
Tokens & tool calls
Every token returned, every tool invocation, including parameters and response bodies.
Decisions & rationale
Which policy applied, which fields were redacted, why a model was selected.
Approvals & reviews
Who approved, when, on what evidence. Dual-control witnesses. Override rationale.
Built to survive an audit.
- Hash-chainedEach event references the previous event's hash. Tampering breaks the chain — provably.
- Signed at the sourceEvents signed by the runtime's KMS key. Public key published in your Trust Center.
- WORM-style retentionWrite-once, read-many storage. Configurable retention from days to years for FINRA, GDPR Article 30, and the EU AI Act.
- Exportable in one clickJSON, CSV, PDF. Per workflow, per agent, per time range. Each export includes a signed manifest your regulator can verify offline.
- Verifiable without usDesigned so your auditor can replay any run using only the export and our public key. No vendor lock-in on evidence.
Send the trace where it needs to go.
Native OpenTelemetry + a custom webhook for anything not on this list.
See it on your stack.
30 minutes with our team. We'll walk you through governance, audit, evals — and answer everything procurement will ask. Bring your own NDA; we'll sign in 24 hours.