Alerts
Agents that are stuck, looping, failed or blocked by policy. Rules run on the controller every 30 seconds.
Open now
5 2 need a hand
2 stuck/failed · 2 denied · 1 loop
Auto-actions today
3 1 pause · 2 retries
opt-in per rule
Time to acknowledge
6m median · 7 d
−4m vs last week
Tokens saved by kills
1.1M 7 d
est. from stuck and loop alerts
Type
What happened
Agent · person
Since
Actions
Stuck
No check-in for 22 min while the pod is alive · same call twenty/mcp/search ×14 with identical args
Rule: stuck after 15 min · auto-kill at 45 min is off for Maria's agents
MD
Haiku WorkerMaria Delgado · "Dedupe CRM leads"
22 minno auto-action
Loop
Same tool, same arguments, 14 times in 5 min · web_fetch https://docs.sulladesktop.com/invite-links · 404 every time
Rule: 10 identical calls in 5 min · auto-pause is on → paused 3 min ago
DP
Codex LunaDev Patel · feat/onboarding-steps
8 minauto-paused
Failed
Workflow node 7 failed · slack/post_message returned 401 · 12 invoices not announced
Rule: run ended in failure · 2 automatic retries used · credential slack/acme-finance may be rotated
PN
invoice-runPriya Nair · webhook trigger
1 min2 retries done
Policy denied
Credential not granted · asked the vault broker for stripe/acme (charge) · item policy allows invoice-run only
Agent carried on without it · 3 denials from this agent in 24 h
MD
crm-cleanerMaria Delgado · v0.3 · her agent
14 minno auto-action
Policy denied
Tool outside scope · called meta/exechost · capability token for server pods has no host scope
Rule: any tool-API denial · blocked by the tool API, nothing ran
JB
Codex SolJonathon Byrdziak · feat/workflows-container
31 minno auto-action
Rules
built in · thresholds per org| Rule | Fires when | Auto-action | Notify | On |
|---|---|---|---|---|
| Stuck | no check-in or event for 15 min while the pod is alive | kill after 45 min more off | badge · Desktop · owner | |
| Loop | same tool + same args 10× in 5 min | pause on | badge · Desktop · owner | |
| Failed | run ends in failure, or a return contract never comes back | retry workflow node ×2 on | badge · Desktop · owner | |
| Denied | any denial at the tool API or the vault broker | — | badge · admins | |
| Pod | OOMKilled, CrashLoop, evicted, disk pressure | — | badge · Desktop · admins | |
| Orphan | a child agent still running 5 min after its parent ended | kill on | badge · owner | |
| Quota | 75% and 90% of a limit | pause at 100% | badge · person |