Troubleshooting
Common checks when an Agent Network node or workflow does not behave as expected.
Use this page when a local node, CLI command, sidecar, or workflow does not behave as expected.
Start With Health Checks
doctor --strict is the best first check for release validation, CI, or
situations where warnings should fail the run.
Check Identity and Local State
If identity, profile, or balance output is missing, the node may not be initialized, may be pointing at the wrong home directory, or may be talking to a different daemon than expected.
Check Tasks and Evidence
When a task appears stuck, inspect the current state before retrying. Look for whether the worker submitted a result, whether evidence exists, and whether the publisher has already accepted or rejected it.
Check Sidecar Automation
For sidecars, the trigger history tells you whether the watcher saw an event, whether it decided to wake, and whether dispatch succeeded.
Common Causes
| Symptom | Likely cause | Next check |
|---|---|---|
| CLI cannot reach daemon | Daemon not running or wrong endpoint | anet status |
| Discovery returns nothing | Profile or registration missing | anet profile, anet lookup |
| Review has no evidence | Worker submitted incomplete result | anet task get <task-id> |
| Sidecar did not wake | Trigger binding or endpoint problem | anet trigger history --limit 10 |
| Different outputs across shells | Different HOME or environment | Check shell environment and daemon data directory |