Change control for AI agents.
Agent memory, approvals, notifications, and audit trail: one API, one signed ledger.
Agents run at full speed, and the trusted record builds itself in real time.
Any agent, any LLM, any platform, over REST or A2A. Self-hosted, even air-gapped.
Agent memory →
Checkpoint work state as signed, typed records. A fresh session of the same agent resumes or takes over in one query, no prior conversation needed.
Approvals →
Routine changes only need recording; the consequential few wait for a verdict. Your own system or a person renders it, and the gate holds the signed answer.
Notifications →
Signed, durable webhooks tell the systems and people around the work the moment an outcome lands. Settlement Signal webhooks ride the same channel.
Audit trail →
Every step signed and hash-chained the moment it happens, into a record nothing can alter afterward. Anyone holding the public keys can verify the chain offline.
Delegation between independent agents
When agents run by different teams, vendors, or platforms delegate tasks to each other, the ask is written once and the remote agent accepts it as written. The API tells each agent what it owes next. An agent starts fresh on every run, so each time it wakes mid-delegation it catches up from the record: who asked, on whose authority, what it already sent out, and whether that is done. In our measurement across three model vendors, a reset agent caught up on the work it had delegated 21 times in 24, against 8 in 23 without the record. Works over REST or A2A 1.0.
AGLedger runs in your own infrastructure: you hold the database, the keys, and the records; it runs air-gapped, with no phone-home. It records what was decided, by whom, and when; the verdicts come from the principal you designate: a person, your own system, or a review agent you run. The free Developer Edition is fully functional and production-capable, one command to deploy.