Loading...
Loading...
OMEGAENGINE · AUDIT LEDGER
Every OmegaEngine decision is tamper-evidently logged with risk scores, safety checks, policy outcomes, hashes, versions, embeddings, and full decision context. Enterprises use this ledger for compliance, audits, legal defensibility, and post-market monitoring under global AI laws.
The Audit Ledger is a cryptographically protected, write-once audit trail for AI reasoning. Every request to /api/authorize, /api/v2/judge, or internal compatibility decision paths generates a structured record.
This is how enterprises prove responsible AI use.
OmegaEngine generates tamper-evident audit fields:
Once a ledger prefix is anchored to the transparency log, it cannot be modified without detection — even by us.
A local ledger produced by the open-source @omegaengine/audit package is tamper-evident to anyone who has the file. The anchor verb makes it verifiable across organizations: it computes the ledger's RFC-6962 Merkle root and submits it to our append-only POST /api/v2/anchor transparency log, which returns an inclusion proof and a signed tree head (STH). This is the one online command in an otherwise offline package.
Run it from CI or a shell — anchoring is available on Pro and above:
export OMEGA_API_KEY=... # your org API key
npx omega-audit anchor ledger.jsonl --api https://omegaengine.aiIt fails closed at every step: an invalid ledger is refused (never anchored), and a rewritten anchored prefix is reported as a fork and rejected — there is no “anchored, probably.” Anchor state lives in a sidecar file next to the ledger (ledger.jsonl.anchor.json), and each subsequent anchor submits an RFC-6962 consistency proof so history cannot silently shrink or rewrite.
What an anchor proves — and what it doesn't
An accepted anchor proves: this exact prefix of this ledger existed, unaltered, no later than the log's STH timestamp. That is the whole claim. It does not prove completeness (entries your agent never wrote are invisible to it) or truth (that the recorded events happened as described). The STH timestamp is operator-attested until external witnessing ships.
The inclusion proof and STH are verifiable offline with the open-source @omegaengine/verify tool — a customer or auditor confirms your record against our public log without an OmegaEngine account, and without trusting us.
OmegaEngine captures one of the richest AI audit logs in the world:
To our knowledge, no other AI decision layer captures this many signals in a single decision record — and you can check: export a record and count the fields.
OmegaEngine fully supports sensitive industries:
You are not selling “an AI agent.” You are selling the **auditability**, **safety**, **compliance**, and **explainability** required for every agent to be used in a real company.
Need help exporting logs for compliance or internal audit?
Contact audit support →