Notary

Turn Agent Chaos Into Court-Ready Evidence

Signed audit records. Passed audits. Notary helps U.S. enterprises deploy AI agents without the compliance cliff.

Book a Demo

30 minutes. We walk you through a live audit export from an actual agent run.

Agent runtimeworkflow-4412
Notary evidencebundle #00241

Your Agents Are Acting. Your Audit Trail Is Vibes.

Ship audit-ready in weeks, not quarters.

Raw signed records, mapped into the formats regulators and auditors accept.

  • EU AI Act Article 12 logs and conformity artifacts
  • NIST AI RMF evidence: Govern, Map, Measure, Manage
  • SOC 2 Common Criteria exports for your AI scope
EU AI ActArticle 12 · Annex IV
  • Automatic event logging
  • Conformity assessment
  • Post-market monitoring
NIST AI RMFGovern · Map · Measure · Manage
  • Control mapping
  • Risk tier evidence
  • Incident records
SOC 2Common Criteria · AI scope
  • CC7.2 system monitoring
  • CC8.1 change mgmt
  • Auditor export

Follow a request across every agent that touched it.

Correlation IDs keep multi-agent handoffs connected, so chain of custody holds all the way to court.

  • Session-level IDs link every planner, researcher, and tool call
  • Graph view of every handoff, reconstructed from signed records
  • Chain-of-custody export aligned to FRE 901 / 902
Session · corr_id 7c21…4a
Chain of custody

Enterprise-grade security, customer-owned keys.

We hold infrastructure. You hold plaintext.

  • SOC 2 Type II certified
  • Customer-managed keys; we never read your evidence
  • VPC / on-premise for regulated industries, MIT-licensed core for self-host
Customer keykms://acme/notary-2026
Regionus-east-1 · VPC

The Evidence Gap, By The Numbers

88%

of enterprises had an AI agent incident last year

21%

could actually prove what happened at runtime

€35M

max EU AI Act penalty for missing Article 12 logs

One evidence schema across every provider you run.

OpenAI, Anthropic, Google, MCP servers, and your agent framework, normalized into a single record shape.

Normalized evidence schemaone shape · every provider

LLM Providers

  • OpenAI
  • Anthropic
  • Google
  • Mistral
  • Azure OpenAI

Agent Frameworks

  • LangChain
  • LangGraph
  • CrewAI
  • OpenAI Agents SDK
  • Claude SDK
  • AutoGen
  • Semantic Kernel
  • PydanticAI

Protocols

  • MCP
  • A2A

Identity & SIEM

  • Okta
  • Entra
  • Auth0
  • Splunk
  • Datadog
  • OpenTelemetry

Frequently Asked Questions

Have a question that’s not here or need help? Contact us.

Ready To Stop Hoping Your Logs Are Enough?

Turn Agent Actions Into Court-Ready Evidence With Notary.

Book a Demo

30 minutes with a founder. No sales pitch.