Operate application callbacks
Configure a callback for the application and environment that own the paused work. VeriProof signs each outcome delivery. The application must verify that signature and the request’s scope before accepting it.
The customer application team owns the listener, durable store, worker, and revision path. Read the Core Human Review lifecycle guide before enabling delivery.
Make the endpoint ready
Confirm the scope
Check the selected application and environment before changing the endpoint.
Let the application load VeriProof’s keys
VeriProof signs each delivery with its own key and publishes the public keys at /.well-known/veriproof-webhook-keys.json on your VeriProof endpoint. The SDK loads them for you. If the application can’t make outbound calls, pin the same key set JSON in its configuration.
Verify the raw request body
Read the request as raw bytes and verify its signature before parsing JSON or taking any action. Reject missing, invalid, stale, or incorrectly scoped signatures.
Durably store the verified outcome, signal the application worker, return the receipt, and stop. Do not resume an agent or run a protected tool in the callback request.
Run the safe test
Send the portal’s safe test and confirm the verified result for the selected endpoint. The test proves signing and delivery without resuming customer work.
Key rotation
VeriProof rotates its signing keys for you. A new key appears in the published key set at least an hour before VeriProof signs with it, and deliveries carry both signatures while the old key is retired. Receivers that load the published key set need no change. If you pinned the key set, update it when VeriProof announces a new key.
Respond to an incident
Start with the incident and delivery shown in the repair panel. Confirm the application and environment, then read the failure reason. Check the receiver’s current health before deciding whether to repair its configuration or replay the delivery.
- Repair corrects the endpoint setup for the selected scope.
- Replay sends the same outcome again with the original delivery identity so the application can handle it safely.
- A durable receipt or execution acknowledgment makes further replay unavailable.
Never disable signature verification, broaden the application scope, or create a new delivery identity to clear an incident. Correct the scoped configuration, then use the authorized repair or replay action.
Callback chains, third-party workflow connectors, and advanced callback fan-out are not available in this release. To keep other tools up to date, send them review event webhooks. Those report changes; they never resume work.