Agents we have built and can deploy.
Grouped by the phase of AIM they run in. Tap any agent to see what it needs, how it is constrained and the number it is measured on. What we can deploy for you depends on your systems and data, so an operator confirms the shortlist before any engagement starts.
Autonomy: Shadow → outputs compared only · Approve → a named person decides · Bounded → acts within limits the risk owner sets
Ten rules no agent we ship may break.
Every output that will be relied on carries a citation to its source.
Decisions against a person — a denial, a decline — are never made by an agent.
Agents start in shadow. Autonomy grows only on evidence, with a signature.
Your data and code stay inside your environment.
Regulated data cannot reach an unapproved model or region.
Every action is logged, attributable and reversible.
Fairness is monitored on any agent that touches people.
No agent ranks or reports on individual employees.
Every token is metered and tied to a project and an outcome.
You can pause any agent in one action, without us.