Skip to content
09

Observe

Audit

Every decision is written to an append-only log where each entry commits to the one before it. Periodic checkpoints are published, so a record cannot be altered or removed after the fact without the chain failing verification.

The problem

Compliance regimes are beginning to ask what autonomous systems did and who authorized it. 'We have application logs' is not an answer that survives a control test.

Capabilities

How Audit works

  • 01

    Append-only, hash-chained

    Each entry includes the hash of its predecessor. Editing or deleting any record breaks verification from that point forward, and the break is detectable.

  • 02

    Independently verifiable

    Checkpoint roots are published on a schedule and can be witnessed externally. An auditor verifies the chain with a tool you do not control.

  • 03

    The record is the decision

    Agent, owner, action, resource, policy, verdict, approver, and inputs — one record answers the question rather than requiring three systems to be joined.

  • 04

    Exports that map to controls

    Scoped exports for SOC 2, ISO 27001, and the EU AI Act's logging expectations, plus streaming to your SIEM in CEF or OTLP.

In practice

What it looks like

Audit recordJSON
{  "seq": 918442,  "ts": "2026-08-06T09:14:02.118Z",  "prev_hash": "sha256:41c9…8ea2",  "hash": "sha256:7bd0…19f4",  "decision_id": "dec_01JQ8F2K7M",  "agent": { "id": "agt_3H8N1P", "owner": "marcus@example.com" },  "action": "k8s.apply",  "resource": "production/checkout",  "verdict": "permit_after_approval",  "policy": "require-approval-prod-deploy",  "approval": {    "approver": "sam@example.com",    "channel": "slack",    "granted_at": "2026-08-06T09:13:44.902Z"  },  "checkpoint": "ckpt_2026-08-06T09:00Z"}

Limits

What this does not do

Stated plainly, because a control plane that claims no limits has either not been deployed or is not being described accurately.
  • A hash chain proves records were not altered after they were written. It cannot prove a record was written truthfully in the first place — that property comes from the enforcement path, not from the log.
  • Tamper-evidence is not tamper-proofing. Someone with sufficient access can destroy a log; the chain guarantees that doing so is visible, not that it is impossible.
  • Cirvix does not certify you against any framework. It produces evidence; an auditor reaches the conclusion.

Get started

Let the agents do real work.

Start with the free local scanner — it needs no account and tells you what is already ungoverned on your machine. Move to a governed runtime when you are ready.

Cirvix AgentControl is in private beta. Pricing is published and plans are not yet chargeable — early access confirms your place and we will tell you before anything bills.