WhatsApp automation not replying? Run statuses and skip reasons
Open the automation's Runs tab. A Skipped run names its reason: "You answered this chat from your phone recently" is the pause after your own reply, 4 hours by default, and "More than 24 hours since the customer's last message" means only a template can go out. A Failed run shows the failing step and its error. No run at all means the message never reached the automation.
easyCoexistence keeps a 30-day log of the automation runs on your WhatsApp number, with each step's output, timing and AI tokens, and emails you when a run fails so you can answer that customer yourself.
A run is Queued, Running, Waiting, Done, Failed or Skipped.
Run skip reasons: a recent reply from your phone, an AI key the provider refused, a chat you paused, an unanswered question.
How far back the Runs tab goes, newest first, with up to 100 runs listed.
At most one failure email per automation in this span; the run history keeps every failure.
Try it runs allowed per hour in the dashboard, with samples of up to 2,000 characters.
Is there a run at all?
Start here, because the two cases have different causes. A run in the Runs tab means the automation picked up the message, and its status says what happened next. No run means the message never reached this automation. The usual reasons, most common first:
- The number is set to "Straight to your system": the list shows "Number off", and messages go straight to your own endpoint. Choosing "Through easyCoexistence" under "Where incoming messages go" on the number's page fixes it.
- The automation is Off, or shows "Paused: AI key refused". Neither starts runs for new messages, and a lower-priority automation may answer in its place.
- The trigger did not match. Type the customer's exact words in the "Test it" box under Keywords.
- Another automation answered first. Keyword rules come first, then First message, After hours and Every message, so check the other automations' runs.
- First message already fired for that chat. It fires once, and again only after 30 days without a message from that customer.
- It was a reaction, a status update or a group message. None of these start runs.
- The number is suspended or disconnected. Check its status on the Connections tab.
What does each run status mean?
The pill on each row of the Runs tab is one of six statuses. Only Failed sends you an email: Skipped means a built-in guard stopped the run on purpose, and Waiting means it is parked on a Wait step or on a question that waits for the customer's answer. Test runs from Try it carry a Test tag and never send anything, so they cannot explain why a customer got no reply.
What does each skip reason mean, and what fixes it?
A Skipped run shows one of 15 labels, and the same label appears on the step where the run stopped. The table below has the nine you meet most, worded exactly. The others are "The bot is paused on this chat" (a Hand to a person step, or a pause set in Conversations or by update_contact), "Too many automatic messages to this chat in an hour (two bots answering each other?)", "They did not answer the question within 24 hours" (an Ask a question step), "Jumped 20 times in this run: the flow went on below instead", "The step to jump to no longer exists", and two that only social accounts use. Most of them are the system doing its job, so the fix is often a setting rather than a bug. Read the label first, then the matching row. Through the MCP connector the same reasons come back as codes, such as human_active, window_closed, ai_limit and superseded.
| What happened | What to do | |
|---|---|---|
| You answered this chat from your phone recently | You replied from the WhatsApp Business app within the pause, 4 hours by default, or during a Wait. | Nothing, if intended. Change "Pause on a chat after I answer it (hours)" under Advanced; 0 turns the pause off. |
| More than 24 hours since the customer's last message | Meta accepts text from the API only within 24 hours of the customer's last message. | Use a Send a template step for later messages, or shorten the waits before a text step. |
| AI reply limit for this chat reached this hour | The chat hit "AI replies per chat per hour, at most", 10 by default. | Raise it under Advanced, up to 60, if real customers reach it; the cap also stops two bots answering each other. |
| A newer message from the customer took over | A newer message started the same automation on that chat and replaced the run that was waiting. | Nothing: the newer run answers. Moving the automation to another number ends waiting runs the same way. |
| Automations are off for this number | While the run waited, the number moved to "Straight to your system", or was suspended or removed. | Choose "Through easyCoexistence" on the number's page, or reconnect the number. |
| The account needs reconnecting | Not used for WhatsApp numbers; a number in that state shows the reason above instead. | Check the number on the Connections tab. |
| Automation is off | The automation was switched Off while this run waited. | Switch it On and press Save. |
| Paused until the AI key works again | The AI provider refused the key, which pauses the automations that use it. | Save a working key under "AI keys"; the automations restart by themselves. |
| Automation was deleted | Rare: the automation disappeared while one of its runs was in progress. | Create it again. Deleting an automation also deletes its run history. |
Why did a run fail?
Open the run and find the step marked Failed: its error is the reason, and each step after it shows Skipped. You also get an email whose subject reads "<automation> could not answer <customer>", with an "Open the run" button, so you can answer that customer from the phone. The errors you are most likely to see:
- "Webhook did not answer": your URL took longer than 20 seconds, or refused the connection.
- "Webhook answered HTTP 500", or another code: only a 2xx answer counts, and redirects are not followed, so a 301 or 302 fails too.
- "Webhook URL refused": the URL is not HTTPS, or it points to a private address.
- "The model returned an empty reply", or an error from the AI provider such as a wrong model id or a rate limit. The AI call gets 45 seconds.
- "Meta (400): ...": Meta refused the send, for example a template that is not approved or a variable left empty.
Why does Try it work when real messages do not?
Because a test checks the steps, not the path a live message takes to reach them. The Test button opens "Try it", and "Run test" runs the saved steps on your sample: the AI really answers on your key and your webhook really receives a POST marked "dry_run": true, but nothing is sent to anyone. It does not check the trigger, the number's mode, which automation would win, your phone pause or the 24-hour window. The AI reads only the sample, without earlier messages, and a Wait shows "Would wait 120 min" and moves on. A passing test proves the steps work; the Runs tab shows what happened to real messages.
How do you read a run?
The Runs tab lists the last 30 days, newest first: the status, a Test tag on tests, the customer's message or a placeholder such as "[image]", their name or number, and when it happened. Open a run to see the skip reason, if any, then each step with OK, Failed or Skipped, the seconds it took, the AI tokens (shown as "312 in, 58 out tokens") and the exact text sent or the error. Your AI can read the same log: through the MCP connector, list_automation_runs filters runs by status and get_automation_run returns one, so asking ChatGPT or Claude why the automation stayed silent this morning gets an answer built from the skip reasons.
How to set it up
- Open the Runs tabOn the Automations tab, open the automation, then Runs. Find the customer's message by its text and time.
- Read the status and the reasonSkipped shows the reason in plain words; Failed shows the step that broke and its error.
- No run: check the number's modeOn the number's page, the card "Where incoming messages go" must be set to "Through easyCoexistence", not "Straight to your system".
- Check the switch and the keyThe automation must be On, not "Paused: AI key refused", and "AI keys" must show the key as Works rather than Refused.
- Check the trigger and the orderTry the customer's words in "Test it" under Keywords, and remember that keyword automations answer before the others.
- Check for your own recent replyIf you answered that chat from the phone in the last few hours, the pause is doing its job; once it ends, the customer's next message gets an answer.
Common mistakes
- Testing from your own phone, then replying from the business app to check. That reply pauses the automation on the chat for 4 hours, so the next tests show "You answered this chat from your phone recently".
- Sending the same test twice to a First message automation. It fires once per chat, so the second message gets nothing; test the steps with "Run test" or write from another phone.
- Taking a passing Try it as proof that live messages will run. It skips the trigger, the number's mode and the 24-hour window, so check the keyword and the number as well.
- Pointing Call my webhook at a URL that redirects. Redirects are not followed, so a 301 or 302 fails the step and blocks the steps after it; use the final HTTPS address.
Your AI can read the same log through the MCP connector and explain it in plain words. US$ 9 down to US$ 0.75 per number per month, falling as you add numbers, with the first 3 days free.
easyCoexistence is a multichannel API and MCP server that connects AI agents such as Claude and ChatGPT to WhatsApp, Instagram, TikTok, Facebook and more. On WhatsApp it uses Coexistence: the business keeps using its number in the WhatsApp Business app while the AI works through Meta's official API.
Frequently asked questions
Why did a customer get two replies?
Usually two systems are answering. The WhatsApp Business app's own greeting or away message keeps working after you connect, and software cannot see it; a bot of your own behind a forward URL can answer too. One automation with both a Reply with text and a Reply with AI step also sends two messages.
Does an automation answer messages that arrived while it was off?
No. A message starts a run only if a matching automation is On when the message arrives, and nothing is queued for later. Answer those customers from the app, or ask your AI to message them within 24 hours of their last message.
How do I stop the automation on one chat only?
Reply on that chat from the WhatsApp Business app. The automation then stays quiet there for the hours set in "Pause on a chat after I answer it (hours)", up to 168, which is a week. You can also switch automatic replies off on one person in Conversations, for 1, 4 or 24 hours or until you turn them back on, or ask your AI to do it with update_contact (pause_hours). Other chats keep their automated replies.
Do failed runs retry on their own?
No. A failed run stays failed, and the email exists so that a person answers that customer. Fix the cause and run a test; the next customer message starts a fresh run. Meta's own retries of a message never create a second run.
Keep reading
Ready to get started?
Set up WhatsApp Coexistence in minutes, not months. The app keeps working on the phone.
Start Free TrialVerified on