Inspect Sentinel

Monitors and control protocols that watch agents and intervene in their execution.

A monitor observes a step of an agent's execution and reports a suspicion score; it never acts. A protocol decides what happens at that step, optionally after consulting monitors, and a protocol with no monitors is a rule. Both run before a tool call executes or after its result comes back; stages around model generation are planned. The same monitor runs inside an Inspect eval and offline over recorded transcripts with Inspect Scout for validation and calibration; running it in a network proxy in front of a model API is planned.

pip install inspect-sentinel

It requires inspect-ai 0.3.278 or later.

The package is under active design. The design documents are in design/; start with sentinel-overview.md.

Runnable examples are in examples/.

Documentation: https://meridianlabs-ai.github.io/inspect_sentinel.