# Eden Control

> Governance today. A safer tomorrow. Eden Control records who authorised each action an AI agent takes through it, checks the action against that authority before it runs, and holds or refuses the rest. It is a working prototype that governs its founder's own AI agents today.

AI proposes. Authority decides. Evidence survives. Every consequential AI action needs an authority envelope (principal, agent, action, purpose, budget, expiry) before it can execute. Each check gives one of three answers: Allow; Investigate, routed to a named human reviewer with an expiry and a fail-closed timeout; or Deny. Every outcome produces a receipt.

## Status (stated plainly)
- Implemented: the operator shell (tenant context, routed gateway, Google identity verification at the API boundary) and the ternary gate engine (a register of 1,244 written rules counted on 14 October 2026; about 176 automatic checks run before an agent's action; 1,514,217 decisions from 5 July to 14 October 2026, 4,877 held for a person and 968 refused).
- Under test: the portable authority envelope with offline-verifiable outcome receipts; external assurance (independent penetration testing, published bypass-resistance evidence).
- No claim of regulator approval or external certification.

## Origin
Built by Michael Darius Eastwood (https://www.michaeldariuseastwood.com/) to govern the AI agents that were building Eden Legal AI (https://edenlegal.ai/). The same system runs the checks of Eden Research (https://www.edenresearch.ai/).

## Links
- Home: https://edencontrol.com/
- Console: https://app.edencontrol.com/
- Verify the record yourself: https://www.michaeldariuseastwood.com/verify/