Explore controls for AI-assisted work
Examine proposed boundaries for agent activity, review, and audit. This interface is experimental; its controls require independent validation before real-world deployment.
Questions to validate before deployment
The dashboard illustrates controls; it does not establish regulatory or clinical readiness.
Parallel Agent Worktrees
Explore isolated git worktrees as a proposed boundary for agent development.
Guardian Merge Gate
Prototype scoring and review controls for proposed code changes.
Live Agent Dashboard
Dashboard views for examining agent activity and review status.
Audit Chain Prototype
Explore hash-chained receipts; external completeness and custody still require validation.
Enterprise RBAC
Prototype roles for studying who may propose, review, and approve an action.
Validation Required
No clinical or compliance approval is implied. Do not enter patient information.
How it works
An illustrative agent workflow to inspect and test.
Spawn
Assign a scoped worktree and proposed capabilities to an agent.
Build
Generate a change and gather test and review evidence.
Score
Inspect changes against an explicit review policy and record uncertainty.
Gate
Require an authorized reviewer before a consequential transition.
Observe
Record what actually happened after an authorized action.
Explore the prototype
Inspect the dashboard and test the proposed controls. Do not enter patient information.
Launch Dashboard