Troubleshooting Runs
When an automation does not do what someone expected, Qlar always leaves a reason: on the automation itself (its status reason) and on each run (its outcome and error code). This page explains every one and what to do about it.
The examples use the automation that Rina Wijaya, Kirana's operations lead, created in Atlas's chat: KG-2041 delay / on-hold check, which runs every weekday at 07:30 (Asia/Jakarta) and delivers a summary only if project KG-2041 is delayed or on hold.
You see them in the Monitor (Automations → Monitor → select a row) and users see the same reasons, in friendlier words, on their Automations page.
Before you start
- You can open the Monitor. See Monitor and Govern.
- You know the statuses and run outcomes described in How Automation Works.
First questions
| Symptom | Look at |
|---|---|
| "It didn't run." | The automation's status. Only Active automations run. Then the Next run time, and whether the policy switch is on. |
| "It ran but I got nothing." | The run's Condition. Condition not met is normal. On a day when KG-2041 is on track, Rina's check ends this way and creates no conversation. Then the Actions column for a failed or skipped action. |
| "I got a summary although the project is on track." | The condition could not be checked, for example because the project status was unavailable. In that case the result is delivered anyway, as the automation preview said. After three such runs in a row, the user gets a Warning notification. Check that the tool or data source the agent uses for project status works. |
| "It ran late." | Runs can start a minute or more after their time, and later when many are due at once. Consider scheduling a few minutes earlier. |
| "It stopped by itself." | The status reason. See the first table below. |
| "The agent says email is not available." | Whether Send email is ticked under Allowed Actions on the policy. A policy saved before email existed does not have it until you tick it and save. If it is ticked, email sending may not be available on the platform right now. |
| "The agent refused my email automation." | The reason it gave. Most often the address is not one valid address, or the user's automations together could send more than 10 emails an hour to that address; a less frequent schedule fixes the latter. |
| "The agent won't offer to create one." | The user must be signed in, outside the CMS preview, and the policy must be on with Create automations allowed. The CMS preview never offers automation, so test in the agent's public chat. |
Why an automation is in its status
| Reason (Monitor) | Status | What it means | What to do |
|---|---|---|---|
| Paused by its creator | Paused | The user paused it. | Nothing. They can resume it. |
| Paused by a contributor | Paused | Someone on your team paused it in the Monitor. | Resume it from the Monitor when ready. |
| Disabled by a contributor | Turned off | Someone on your team disabled it. The user cannot resume or edit it. | Resume it from the Monitor, or the user can delete it. |
| Paused: more active automations than the policy allows | Paused | You lowered Maximum Active Automations. The newest above the limit were paused. | Raise the limit, or ask the user to delete some. |
| Paused: the agent moved to another organization | Paused | The agent was transferred. | Users resume what they still want, after checking it. |
| Stopped after repeated failures | Stopped | Three runs failed in a row. | Check the history for the cause (below), fix it, then resume. |
| Stopped: runs more often than the policy allows | Stopped | You raised Minimum Interval above its schedule. | The user edits the schedule, then resumes. |
| Stopped: its trigger is no longer allowed | Stopped | The trigger kind was removed from Allowed Triggers. | Allow it again, or the user recreates it. |
| Stopped: none of its actions is allowed any more | Stopped | All its actions were removed from Allowed Actions. | Allow an action again, or the user edits it. |
| Stopped: its creator lost access to the agent | Stopped | The user left the organization or lost access to the agent. | Restore the access, or delete the automation. |
| Stopped: its content could not be read | Stopped | The encrypted content is unavailable (Privacy Mode). | Ask the user to recreate it. |
| Its schedule has finished | Finished | A one-time run happened, or the end date passed. | Nothing. Removed after 7 days. |
| Reached the policy's limit of runs | Finished | Maximum Runs per Automation was reached. | Raise the limit, or the user creates a new one. |
Once you have fixed the cause, use Resume in the Monitor (if you may manage automations) or ask the user to resume from their page. Resuming checks the policy again, so it will refuse if the cause is still there.
Why a run failed or was skipped
Each run in the history shows one of these codes.
Policy and limits
| Code | Meaning | What to do |
|---|---|---|
| Automations are switched off in the policy | The master switch is off. The run is skipped, not lost. | Switch it back on. |
| Trigger not allowed by the policy / No action allowed by the policy | Your policy no longer allows it. | Adjust the policy. |
| Runs more often than the policy allows | The interval is below the Minimum Interval. | The user edits the schedule. |
| User's daily run quota reached | Maximum Runs per User per Day is used up. | Raise it, or ask the user to reduce how often theirs run. |
| Run limit reached | Maximum Runs per Automation reached. | Raise the limit. |
Credit and budget
| Code | Meaning | What to do |
|---|---|---|
| Organization credit exhausted | There is not enough credit to run. | Buy credits or set up auto top-up. |
| Member budget or agent limit reached | The user's budget for this agent, or the agent's own limit, is used up. | Raise the budget or the agent limit. |
The run itself
| Code | Meaning | What to do |
|---|---|---|
| Timed out | The run took longer than Run Timeout. | Raise the timeout, or simplify the task or condition. |
| The agent's run failed | The agent could not finish. Temporary failures are retried automatically. | Check the Monitoring Logs. If your tools are down, repair them. |
| A selected skill is no longer available | A skill the automation used was removed. The run continues without it. | Restore the skill, or the user edits the automation. |
| Content could not be read | The encrypted content is unavailable. | Ask the user to recreate it. |
| Agent not found | The agent was deleted or unpublished. | Restore or republish it. |
| Creator no longer has access | The user lost access to the agent. | Restore access. |
Timing
| Code | Meaning | What to do |
|---|---|---|
| Skipped: started too late | The automation uses Skip as its missed-run policy and could not start on time. | Change its missed-run policy, or wait for the next run. |
| Superseded by a newer schedule | The user edited it and a newer run replaced this one. | Nothing. This is expected. |
| Abandoned | The run was interrupted before it finished. | Nothing, unless it repeats. |
| Paused platform-wide | Qlar paused all automations for a moment. | Nothing. Runs resume. |
| Code | Meaning | What to do |
|---|---|---|
| Invalid WhatsApp number | The number cannot be a real phone number. | The user corrects it. |
| Daily WhatsApp limit reached | Maximum WhatsApp Messages per Day is used up. | Raise the limit. |
| WhatsApp rate limit reached | This number, or the agent's WhatsApp number as a whole, reached its hourly limit for automation messages (10 an hour to one number, 30 an hour in all). Qlar tries again in the next hour. | Nothing. |
| No WhatsApp channel that can send (needs the public or private server) | The agent's Published configuration has no WhatsApp channel on the public or private server, or it is switched off. Meta Cloud, Coster and Qontak can only start a conversation with an approved template. It is not retried. | Connect a number on the public or private server in Channels > WhatsApp, Published configuration, or switch the channel back on. |
| Agent's WhatsApp channel not connected | The agent's WhatsApp number is not linked right now, for example because the phone was logged out. Qlar tries again later. | Check the channel in Channels > WhatsApp and link it again if needed. |
| Number not allowed by the WhatsApp channel's contact rules | The channel is in private mode and this number is not on its list, or the number is blocked. It is not retried. | Allow the number in the channel's contacts, or have the user choose another number. |
| No WhatsApp sender available | Seen only on runs before 7 October 2026, when results were sent from the platform's shared number. | Nothing. |
| WhatsApp rejected the message | WhatsApp refused it. | Check that the number can receive WhatsApp messages. |
| WhatsApp send failed | The message could not be handed to WhatsApp. It is not retried, in case it went out. | Ask the user whether it arrived; check the agent's WhatsApp channel if it repeats. |
| WhatsApp outcome unknown | Qlar cannot tell whether it arrived. It is not retried, to avoid a duplicate. | Ask the user whether it arrived. |
An email is never retried automatically, so it can never arrive twice. Whatever went wrong, the next scheduled run sends as usual.
| Code | Meaning | What to do |
|---|---|---|
Email sending is not available (email_not_configured) | Email sending is not available on the platform at the moment. | Nothing on your side; the other actions of the run still happen. Contact Qlar support if it persists. |
Daily email limit reached (email_daily_cap) | The agent's Maximum Emails per Day, or the platform's own daily limit, is used up. | Raise Maximum Emails per Day on the policy. If the agent's limit is not reached, wait for the next day. |
Email service busy (email_rate_limited) | The mail service was briefly overloaded and did not accept the email. | Nothing. The next run sends normally. |
Email address refused (email_rejected) | The mail service refused the address or the message, for example an address that does not exist. | Check the address with the user; they correct it on their Automations page or by asking the agent. |
Email could not be sent (email_failed) | Sending failed for another reason, such as a network error. It is not retried, in case it went out. | Ask the user whether it arrived, including the spam folder. Contact Qlar support if it repeats. |
Invalid email address (email_recipient_invalid) | The stored address is not one valid address. | The user corrects it. |
Retries
A temporary failure, for example an AI provider that is busy, is retried up to Maximum Retries times, waiting Retry Interval minutes between attempts. Each attempt is a separate line in the history, marked Retry. A WhatsApp message or email is never sent twice, and a run that never got as far as sending it is safe to retry. Each action is handled on its own: an email is never tried again, while a WhatsApp message in the same run that had to wait, for example for the next hour, still goes out later.
Three runs in a row that fail after all retries stop the automation (Stopped after repeated failures). A single success resets the count.
Still stuck?
Ask the user for the automation's name and the time it should have run. For example, Rina says she got nothing on Tuesday morning: look up KG-2041 delay / on-hold check in the Monitor and read the 07:30 run of that day. Condition not met means the project was on track; a failure shows its code from the tables above. If the history shows nothing at all at that time, check that the automation was active and that the policy was on. For anything in this page that you cannot resolve, contact Qlar support with the automation's name and the run time.
Check the result
- After fixing the cause, open the automation in the Monitor (Automations → Monitor, select the row). If it is Stopped or Paused, click Resume. Resuming checks the policy again and refuses if the cause is still there.
- Ask the user to click Run now on their Automations page, or wait for the next scheduled run.
- In Execution history, the new run shows Succeeded, with the condition Met or Not met and, when met, each action as sent. For Rina's check that is Create conversation: Sent.
Next steps
- Next in the journey: Fine-Tune Answer Quality, to improve the answers your agent (and its automations) produce.
- Help → Monitor: every field of the Monitor page.