01
Behavior Analysis
Inspect how an agent reasons and where its behavior originates.
//
AgentDeprogramming
Tools for analyzing and modifying AI agent behavior.
Supporting ToolDARE Labs
AgentDeprogramming is a research toolkit for looking inside AI agent behavior: understanding why an agent acts the way it does, and modifying that behavior when it drifts from intent.
It reflects the same conviction behind Cerebro's Domain Evals: you can't trust a model you can't test and correct.
02
Targeted Modification
Adjust agent behavior without retraining from scratch.
03
Safety Tooling
Catch and correct unwanted behaviors before deployment.