AgentLedger

Evidence layer

A tamper-evident audit trail layer for AI decisions. It answers “can we really let AI handle this?” with a record.

Talk to us (30 min) See SynthAgent

AgentLedger — Audit Trailchain verified

08:32:11 · journal entry created

action=create_deal amount=— status=approval_required

08:33:04 · CEO approval

approved_by=CEO gate=CEO status=approved

08:33:05 · execute + signed log

prev_hash=… sha256=… recorded

Illustrative example. No action is executed before approval.

WHAT IT RECORDS

Who, what, when, and on what basis.

Who

The approver and the acting principal are recorded, keeping human and AI decisions separate.

What

Each action — created, approved, executed — is recorded with its target and result.

When

A timestamp is kept for every action, so you can trace the sequence afterward.

Tamper-evident (hash chain)

Records are chained with prev_hash and sha256, so later rewrites can be detected.

Built into SynthAgent / standalone offering planned

AgentLedger is built in as the governance foundation of SynthAgent. A standalone offering — adding only the audit trail layer to your existing systems — is also planned. Planned for launch on AWS Marketplace in September 2026.

Answer with a record. Start with 30 minutes.

We will help design which evidence to keep for which operations, including the approval and logging design. Inquiries in English are welcome.