Exact-action authority
Bind principal, agent, resource, purpose, budget, expiry and proposed-action hash before execution.
Governed AI infrastructure · Private beta
The boundary between permission and consequence.
Eden Control is being built to prove that an AI action was authorised by the right principal, for the exact purpose, resource, budget and time window—and to preserve what happened afterwards.
Private beta. No claim of regulator approval, completed external certification or autonomous safety.
The control model
AI proposes. Authority decides. Evidence survives.
Contestable Investigate
Route uncertainty to named reviewers with evidence, expiry, dissent and a fail-closed timeout.
Portable outcome proof
Connect authority, decision, approval, execution and observed outcome in a record designed for independent review.
Current proof posture
Clear status beats inflated claims.
Governed operator shell
Separate Control shell, explicit tenant context, routed gateway domain and approval-aware swarm actions.
Supabase session gate
Google owner identity is checked at the browser and governed Netlify API boundary before Control access.
Portable authority envelope
Exact-action delegation and offline-verifiable outcome receipts remain a bounded development programme.
External assurance
Independent penetration testing, customer deployments and published bypass-resistance evidence remain release gates.
Private beta
Built first as Eden's own production boundary.
The current console governs internal operations while the authority model is hardened. Access is restricted to the configured owner identity.