Capability · Audit
An audit trail you can verify, live
Every proposal, decision, and dispatch on a signed, tamper-evident record — with a live view of what is running right now.
Delegation without records is just hope. monopea writes everything down: every proposal the agent drafts, every approval or rejection you make, every dispatch that follows — a complete account of what the agent did, what it asked to do, and what you decided.
The record is built to be believed. Audit-chain checkpoints are Ed25519-signed, so the history is tamper-evident rather than merely stored — and the same data feeds a live activity view, so the trail is something you watch as it forms, not just something you consult afterwards.
What gets recorded
The audit trail captures the full decision lifecycle: the proposal with its exact tool and arguments, who approved or rejected it and when, the dispatch and its result, and the policy grants that let routine calls flow. If it touched the world — or asked to — it is in the ledger.
Because the gate and the log are the same pipeline, the trail is not a best-effort mirror of what happened; it is the mechanism through which things happen. Every dispatch is a logged dispatch.
Tamper-evident, not just stored
Checkpoints over the audit chain are signed with Ed25519, which means the history can be verified, not just read: altering or deleting an entry after the fact breaks the chain visibly. The difference between a log and evidence is exactly this property — a claim you can check instead of one you must trust.
For an operator answering to a client, a regulator, or their own future self, that changes what the trail is worth: “here is what the agent did” becomes a checkable statement.
A live view of right now
The same events stream into a live activity view: what is running, what is queued, what is blocked on a decision, and the reasoning behind each move. You can watch a run take shape in real time instead of reconstructing it tomorrow — and act on what you see while it still matters.
That live legibility is what makes it comfortable to hand over real work, and it pairs with the controls: see something drifting, steer it with a message; see something wrong, stop it.
- Running, queued, and blocked work — visible as it happens
- The reasoning behind each move, on the record
- Steer or stop directly from what you observe
Records built for oversight
Human oversight is only as strong as its evidence. monopea’s trail documents the oversight itself — proposals held, decisions made, grants issued — which is precisely the kind of external, demonstrable control the EU AI Act’s Article 14 oversight obligations contemplate. Built for that role; not a certification claim.
And the trail is yours: it lives with the rest of your tenant data, stored in Zurich, under the same isolation guarantees as everything else.
Why it matters
Nothing happens in the dark
Every proposal, decision, and dispatch is on the record — with reasoning attached.
Verifiable history
Ed25519-signed checkpoints make the audit chain tamper-evident, not just persistent.
Live, not retrospective
Watch runs, queues, and blocked decisions in real time — and act on what you see.
FAQ
Audit trail & live activity, in short
- What exactly is logged?
- The full decision lifecycle: proposals with their exact tool calls and arguments, approvals and rejections, dispatches and their results, and the policy grants that allow routine calls to flow.
- Can the log be tampered with?
- The audit chain is checkpointed with Ed25519 signatures, so alterations break the chain detectably. The history is tamper-evident — designed to be verified, not merely trusted.
- Can I see what the agent is doing right now?
- Yes. A live activity view shows what is running, queued, and blocked on a decision, with the reasoning behind each move — and you can steer or stop from there.
- Does this help with EU AI Act obligations?
- It is built for that role: the trail documents the human-oversight control itself — what was held, who decided, what was granted. That is architectural support for Article 14-style oversight, not a compliance certification.
Keep exploring
Approval gate
Default-deny control: mutation-capable tool calls become pending proposals and wait for your decision before anything is dispatched.
Goals & autonomous runs
Hand the agent an objective and it plans and runs the work — on schedules, around the clock — while you steer, stop, and approve.
Delegation & sub-agents
Big jobs fan out to sub-agents that run in parallel — every one of them inheriting the same approval gate, and all of them stoppable at once.