Review your first production incident

Availability: preview

Use the inbound URL returned for the saved signed webhook. Treat the token in that URL as a credential.

1. Send a signal

curl -X POST \
  -H 'Content-Type: application/json' \
  -d '{"title":"checkout error rate increased","service":"checkout-api","severity":"P2"}' \
  'http://127.0.0.1:8080/hooks/REPLACE_WITH_INBOUND_TOKEN'

A valid request returns 202 {"accepted":true}. This confirms synchronous acceptance only. Normalization, Observation persistence, incident ingestion, and any eligible remediation start continue asynchronously in the API process. Model failure falls back to deterministic grouping rather than dropping a valid webhook.

2. Find the incident

curl -b /tmp/mendry-cookie.txt \
  http://127.0.0.1:8080/api/v1/projects/checkout-api/incidents

Repeated signals with the same fingerprint update the open incident occurrence. Closed or recovered incidents are not reopened by webhook ingestion.

3. Review evidence and remediation

Open the incident in the console, or request its remediation review using the public incident identifier returned by the list:

curl -b /tmp/mendry-cookie.txt \
  http://127.0.0.1:8080/api/v1/projects/checkout-api/incidents/INC-REPLACE/remediation

Confirm that facts cite stored evidence, unavailable sources remain explicit, and candidate changes stay separated from human merge and deployment. A review may report insufficient evidence when repository or source capabilities are not configured; that is a valid bounded result, not permission to bypass the gate.

Use troubleshooting if the request is accepted but no incident appears.

Verified against

• backend/README.md