EXPERIMENT / INSPECT / COMPARE

Can the next agent continue?

Change an observation. Inspect the consequence. Keep the evidence.

Validated example contracts · fresh browser checks
Not a live agent run or an authorization service.

Continuation decision

READY

Load a scenario

Preparing the supplied example contracts.

Expected ↔ observed

Your evidence comparison will appear here.

Evidence package

Inspect exact artifacts

Export includes exact input bytes and hashes. Matching hashes prove byte identity, not authenticity. Full contract validation remains with the Python CLI.

EXPERIMENT NOTEBOOK

Compare what changed

No runs yet

Evaluated runs stay in this tab only. Export any result before leaving.

BaselineCompare with

Evaluate twice to compare decisions and evidence.

CONTRACT REGRESSION STUDIO

Exercise the failure boundary

Simulated cases test mechanical decisions, not model quality or production reliability.

Reopen an experiment

Import a workbench JSON bundle

Read locally, never uploaded. Only bundles bound to the supplied example contracts are supported; arbitrary handoffs still use the CLI.