What it does
Provide a controlled agent-action workbench with structural perception, default-deny grants, bounded effectors, re-perception, rollback, and a durable journal.
- bounded agent actions
- rollback-aware automation
- action receipt review
Who it is for
Audiences
- authorized security teams
- evaluation engineers
- agent-system engineers
Deployment contexts
- owned or expressly authorized systems
- review and CI lanes
- controlled agent workflows
Representative workflow
Inputs
- bounded agent actions inputs
- reviewer-defined constraints
Outputs
- bounded agent actions result
- reviewable verification record
Run or evaluate
Entry
python -m pip install -e ".[test]"; python examples/demo.pyVerification
python -m pytestArchitecture and relationships
Dependencies
None declared in the public registry.
Accountable Machines is the companion theory route, not this product route.
Current evidence
verified · 2026-08-27
Limitations and authorization boundaries
- The model cannot supply its own authorization.
- Self-authored tests do not establish independent adversarial review or production safety.
Authorization boundary. Every effect requires authority external to the model and remains limited to the granted scope.