Mission control
for AI agents.

Banks, hospitals, and claims offices are putting AI agents in the loop on decisions that used to take a human. Most of the time the agent is fine. Sometimes it produces a wrong answer that was perfectly internally consistent: every reasoning step was reasonable, every tool call was the one it usually makes, and yet the outcome was wrong, for a reason that is hard to see after the fact.

When that wrong answer surfaces months later in an audit or a regulator examination, the trace is usually gone, the model has been re-deployed twice, and the engineer who configured it left the firm. What is missing is a way to reconstruct the decision and stand behind it, long after the moment that produced it has passed.

The lab's work centers on one question: how do you keep AI systems inspectable, safe, and accountable when they ship to real production work? That agenda builds on a decade of anomaly and outlier detection at scale. PyOD is the clearest deployed example: a practical library used across fraud detection, intrusion analysis, operations, and scientific monitoring. Auditing the decisions AI agents make is where that discipline goes next. When an agent's call becomes part of the production system, the same posture applies.

For an agent's decision, the question is whether you can stand behind the call later, when the context it was made in is gone and someone asks you to account for it. The engineering posture that made PyOD trusted (reproducibility, scale, evidence quality) carries over directly.

auditable recovers an agent run as one typed graph and reads it at three points. Before deploy, it lints a declared plan for the structural risks a reviewer would miss. While the agent runs, it captures the dependency state a decision relied on, replays the decision against the state that is live now, and reverses the committed action through a rail when it no longer holds. After a run, it ranks the finished trace and names the one step the rest of the run rests on.

Your logs show what the agent did. auditable shows what it relied on. Most tools record the call; auditable re-decides under live state and rolls the stale action back.

The graph is measured against real agent runs. In GRADE, across six public agent corpora, the dependency layer predicts which runs fail at ROC-AUC 0.805 where run length carries no signal, and the execution layer localizes the faulting step at Top-3 0.614.