Every action leaves a sealed record.
When somebody asks who did what, six months later, you do not have to go and reconstruct it. Every click, every tool an AI agent used, every model answer and every decision lands on one record as it happens, and the record is sealed, so a later edit to it would show.
Every line is assigned to whoever acted, and names the human whose authority they used.
One record, everybody on it
A human's clicks, an AI agent's tool use, a model's answer, an approval and a refusal all write to the same place. When an AI agent acts for a human, the line is assigned to both: the machine that did it, and the human whose authority it was acting on.
- Who, what, on what, when, and on whose behalf.
- Every tool an AI agent uses is logged with the reason it gave, what came back, and how long it took.
- Lines are added, never rewritten. They are removed only on the schedule you set.
Sealed, so a later edit shows
Every so often, a window of lines is locked together with a fingerprint that includes the fingerprint of the window before it. Ask the platform to check, and it recomputes every fingerprint from the lines it still holds and tells you whether they still match. It shows tampering. It is not a safe, and it does not pretend to be one.
Evidence you export, not evidence you write
For each project the platform works out where each of your controls stands: met, gone stale, missing, or not something this project has to answer for. It packs the answer into one file you can download, worst first. What it cannot see for itself, it assigns to a human in your organisation to vouch for, rather than quietly counting it as done.
A parts list for the software
Every project gets a versioned list of what its software is built from. Read it as a document, or export it in the format the industry already exchanges.
What we do not claim
The platform is not certified to any framework; certificates are awarded to the way an organisation runs itself, never to a piece of software. What it gives you is the evidence your own management system would bring to that conversation.
Ask which of your controls we can evidence from the record.
Also in Governance
- ApprovalsAnything that cannot be undone is assigned to a human.Read the page
- AI agent securityA name, a limit, and somebody who can stop it.Read the page
- AI agent identityA human lets an AI agent in, and can stop it again.Read the page
- What an AI agent may doIt never reaches further than the human it works for.Read the page
- AI agent sign-insA sign-in expires sooner the more it can do.Read the page
- Approved modelsWhich models may run, who approved each one, and what it left behind.Read the page
- Scans and approved packagesApproved packages at the exact version, checked when used.Read the page