Workflows can fail silently.
Yours won't.
Start a canary the moment a workflow kicks off. Resolve it the moment it finishes. If your workflow times out, CrashGuard alerts you the moment it happens.
Different from log-based monitoring.
Traditional monitoring watches what already happened, after the fact, and asks you to guess what a failure looks like in a sea of log lines and metrics. CrashGuard is different.
CrashGuard
- Proactive — you assert success up front, CrashGuard watches for its absence
- One canary per workflow instance, regardless of how many systems it touches
- Alerts the instant a deadline passes, with no anomaly detection guesswork
- Confirms the workflow-specific definition of "done" — not just "it logged something"
Three steps. No black box.
Create
Start a canary from anywhere with the C#, Go, or Rust SDK when an asynchronous workflow kicks off.
Resolve
Mark the canary resolved from anywhere else once the workflow has completed successfully.
Get Alerted
If a canary doesn't resolve in time, CrashGuard flags it immediately so you can investigate.
Built for real workflows, not generic heartbeats
Custom verifiers
Resolving a canary doesn't have to mean "trust the caller." Attach a custom verifier to any canary type, and CrashGuard will call out to your workflow-specific logic before marking it as having failed — checking that an order total matches, a payment cleared for the right amount, or a file landed in the expected location. If the verifier rejects it, you get alerted even though the workflow technically reported in.
Conditional alerting rules
Define rules against canary fields or metadata — route a high-value order's expiry straight to PagerDuty while a routine batch job just logs a warning. Rules evaluate against the same metadata you attached when the canary was created, no separate config duplication required.
Arbitrary metadata
Attach any JSON payload when you create a canary. When a canary is triggered you'll get all this detail returned to you through the alerting channels you configure.
Self-hosted by default
The engine runs in your own infrastructure, backed by SQLite out of the box. Your workflow data never has to leave your network to be monitored.
Stop finding out from your customers.
Drop the SDK into your workflow, point it at your own engine, and let CrashGuard watch your back.