Set up a WhatsApp automation in WACRM and read its run log

An automation in WACRM is a rule: when something happens — a new order, an order marked fulfilled, an abandoned cart, an inbound message, or a booking reminder — it runs one or more actions against the contact that event is about. This tutorial walks the real Automations screen, the 5 triggers and 4 actions that exist today, how to turn a rule on or off, how to read its run log, and the same thing through the API.

What you need before you start

• A WACRM Workspace account with at least one project.
• For the New order, Order fulfilled and Cart abandoned triggers: a connected store — these fire from real store activity, not from a demo. Whether your plan includes a store connection is not listed on the pricing page beyond “Store connector (early access)” on the Advanced plan; check your plan in Settings if you’re unsure.
• For the Send a template action: at least one approved template, since the action only lets you pick from approved templates.
• For the API steps only: curl (or any HTTP client that can set a header), and an API key from Developer / API & MCP.

Create an automation in Workspace

1. Open Automations in Workspace.
2. Click New automation.
3. Fill in a Name, then pick a Trigger. If you pick Booking reminder, a Reminder timing field appears — choose 24 hours before or 2 hours before; the rule only runs for that one timing, never the other.
4. Under Actions, click Add action for each action you want this rule to run when it fires. Pick an Action type and fill in that type’s fields (see the table below). Leaving Actions empty still saves a rule — it fires and gets recorded in the run log, it just doesn’t do anything.
5. Click Create automation.

The 5 triggers

New order and Order fulfilled typically back order-confirmation and status-update sends, the kind of message WhatsApp classifies as a Utility conversation, not Marketing.

The 4 actions

An action with no contact to work with (no contact resolved from the triggering event) is skipped, not failed — Add a label and Create a ticket both need one. Propose an agent action is also skipped while the workspace’s AI is paused. Send a template checks contact, AI pause, consent and the template’s approval status, in that order, before it sends anything — the same Approved status Meta’s own template creation flow in WhatsApp Manager requires before a template can be used at all. See “Read the run log” below for what each of those looks like when a rule doesn’t send.

Turn a rule on or off

Each row in the automations list has an Enabled switch, shown as On or Off. Flipping it saves immediately; if the save fails, the switch reverts to its previous state and an inline error appears — nothing changes silently. Disabling a rule stops it from firing on new events; it doesn’t touch runs it already recorded.

To remove a rule entirely, use the row’s Delete action. A confirmation dialog asks you to confirm: “<name>” will be permanently removed. This cannot be undone.

Read the run log

Each row has a Last runs column with a View last runs toggle. Opening it (only when you click it — the table never fetches runs for every rule just from loading the page) fetches the rule’s 5 most recent runs and shows each one’s status:

A Skipped or Failed run also shows a reason, in plain language: No contact on this event, Contact not found, Label not found, AI is paused for this workspace, Contact hasn’t opted in, Template not found, Template not approved, Send failed, Unknown action type, or Error while running this action. A Send a template action whose send fails for any reason — not just a denied consent check — is always recorded Failed with reason Send failed, never Skipped, so a real delivery problem is never shown the same way as an expected “contact hasn’t opted in” skip.

A run that crashes

A run is claimed the moment its trigger fires and finalized once every action has run. If the process dies in between — a crash, a restart — the run is left Pending and isn’t retried immediately. A scheduled check (today, every 15 minutes) looks for pending runs older than 15 minutes and resumes them from where they left off: any action already recorded in that run isn’t re-run, so a template that already sent is never sent a second time — the same idempotent retry-safety a resumed run relies on. A run gets up to 3 attempts in total; once it’s exhausted them it’s marked Failed, with a reason the run log doesn’t yet have a translated label for — you’ll see the raw value requeue_exhausted instead of a sentence.

The same thing through the API

Automations are also a Core API resource, under /v1/crm-projects/{project}/automations. Reads use the agent:propose scope, writes use agent:approve — tick both under Restrict what this key can do (advanced) → WhatsApp CRM scopes when you create a key in Developer / API & MCP (agent:approve needs a Secret key).

Create a rule with an action

curl -s -X POST "https://api.wacrm.ai/v1/crm-projects/your-project-uuid/automations" \
  -H "Authorization: Bearer sk_live_your_scoped_key" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Tag new orders",
    "trigger": "order_created",
    "enabled": true,
    "config": { "actions": [{ "type": "add_label", "label_id": "your-label-uuid" }] }
  }'

{
  "data": {
    "id": "0e2b1c9a-...-automation-uuid",
    "name": "Tag new orders",
    "trigger": "order_created",
    "config": { "actions": [{ "type": "add_label", "label_id": "your-label-uuid" }] },
    "enabled": true,
    "created_at": "2026-11-08T09:00:00+00:00",
    "updated_at": "2026-11-08T09:00:00+00:00"
  }
}

An unrecognized trigger answers 422 invalid_trigger; a config that doesn’t match what that trigger (or that action type) requires answers 422 invalid_config — e.g. booking_reminder without a kind of 24h/2h, or an add_label action whose label_id isn’t one of your own project’s labels.

List a rule’s runs

curl -s "https://api.wacrm.ai/v1/crm-projects/your-project-uuid/automations/your-automation-uuid/runs?limit=5" \
  -H "Authorization: Bearer sk_live_your_scoped_key"

{
  "data": [
    { "event_key": "order:9f2a...", "status": "ok", "reason": null },
    { "event_key": "order:7c10...", "status": "skipped", "reason": "no_contact" }
  ]
}

Newest first, same fields the Last runs panel itself reads, limit 1-100 (default 20). A project id your key doesn’t own, or an automation id from another project, both answer 404 not_found — never 403, so you can’t tell the two apart from the response. A key missing the required scope answers 403 insufficient_scope.

What works today vs. what’s not there yet

Ready to put this into practice? Request early access to get a Workspace, or have WACRM’s team set up your project for you with Quick Setup.