Skip to main content
Use the Darwin status page for current service availability and incident communication.

​
During an incident

  • Preserve the user’s query or Action ID.
  • Retry only transient failures with bounded backoff and jitter.
  • Keep nonterminal work nonterminal.
  • Do not create replacement Actions until you read the existing Action state.
  • Include Darwin’s response correlation ID in support requests, but never include credentials.
Provider-specific unavailability can appear as a structured Action state even when the Darwin API itself is healthy.
Last modified on September 21, 2026