Agent and client identity
Requests bind to trust status, credential envelope, scoped access key, task token, and allowed tool route.
Wever Labs runtime rails carry intake records, Work Orders, provider confirmations, credits, tool commands, result contracts, receipts, callbacks, ledgers, and attestations. The security boundary defines how those records are authenticated, scoped, validated, written, delivered, retried, and disabled when needed.
The OS security layer keeps public discovery separate from runtime writes, service-role actions, provider secrets, callback delivery, and controlled pilot execution.
Requests bind to trust status, credential envelope, scoped access key, task token, and allowed tool route.
Provider credentials, webhook secrets, Supabase service-role keys, and signing material stay in environment variables and server-side functions.
Runtime functions validate tool, schema version, callback target, provider event, credit funding event, and result envelope before writes.
Security-relevant actions create audit events for intake, dispatch, execution, provider confirmation, callback delivery, retry, and kill-switch use.
TokenOps, FinanceOps, and EnergyOps run through the same control model: trusted entry, scoped execution, validated provider events, bounded callbacks, audit events, and operator-visible state.
Bind request identity to agent, client, credential envelope, and allowed tool scope.
Check payload schema, provider event, callback URL, and Work Order route before runtime writes.
Use server-side service-role actions to create runtime records, credit events, and result objects.
Create security events for sensitive actions, retries, failed checks, and operator overrides.
Provider routes, callback delivery, and pilot runs can be disabled by kill-switch state.
Move from the security boundary into runtime deployment, function wiring, callback delivery, or controlled pilot execution.