Skip to main content
A loop tells QA Wolf what to do when something happens. It has two parts:
  • When — what starts the loop: a deploy, a schedule, a pull request, or a QA Wolf event.
  • Then — what QA Wolf does each time: Run AI with your instructions, or Run specific flows.
Loops belong to the workspace, so one loop can cover staging, production, and every preview environment. You find them under Loops in the main navigation. The public API and CLI call a loop a trigger, through trigger.*. See Trigger fields.

When

The Starts on menu picks what starts the loop:
  • Deploys to an environment — fires when your pipeline notifies QA Wolf of a successful deploy to staging, production, or any long-lived environment. See PR testing and the deployment testing guides for how the notification reaches QA Wolf.
  • A schedule — fires in one environment every few hours, days, weeks, or months. Pick a shortcut such as Hourly, Weekdays, or Every 2 weeks, or set Every [N] [unit] at [time]. An hourly or daily schedule can be restricted to chosen days of the week. On an hourly schedule, Only certain hours sets a starting at time and an optional until time, and every run happens at the starting time’s minute. Needs nothing from your pipeline.
  • Pull request previews — fires on deploys to preview environments named like a pattern, such as pr-*.
  • A pull request — fires when a pull request is opened, marked ready for review, gets new commits, gets a label, or is merged. Needs a connected GitHub or GitLab integration. See Pull request events.
  • A bug, A blocker, or A request — fires on a QA Wolf event. See QA Wolf events.
To narrow a deploy loop, select Only when… and add a repository, service, branch, or deploy URL condition. Each condition suggests the values your recent deploys reported, and you can type a name or pattern instead. The editor says how many of your recent deploys the loop would have run on, and lists them.

Pull request events

A loop that starts on A pull request fires on the events you tick:
  • Opened — a pull request is opened.
  • Ready for review — a draft pull request is marked ready for review.
  • Commits pushed — new commits reach the pull request. A burst of pushes fires the loop once, after two minutes with no new push.
  • Labeled — the pull request gets a label. Ticking it adds a Label filter that names the label.
  • Merged — the pull request is merged.
Repository picks the repositories the loop watches, such as acme/web or acme/*. Into branch and From branch narrow it to pull requests between matching branches, and Skip draft pull requests leaves drafts out. Each filter takes a name or a pattern. The editor lists the recent pull requests the loop would have fired on. To test each pull request’s preview deploy instead, select Or, once it’s deployed: Preview deployed. That makes the loop start on Pull request previews. A loop on a pull request can Run AI or Run specific flows. With Run specific flows, the Environment option picks where the flows run, as on a QA Wolf event.

QA Wolf events

A loop can start when something happens to a bug, a blocker, or a request in QA Wolf. Each one has its own events in the Starts on menu: gets a priority asks for the priorities that start the loop, and moves to a status asks for the statuses. A request that moves to In review counts as moving to In Progress. To narrow a loop further, add a Priority filter, or a Blocked on filter for a blocker. A loop on a QA Wolf event can Run AI or Run specific flows. With Run specific flows, the Environment option picks where the flows run. It defaults to the workspace’s default environment. When the environment a loop was saved with is terminated, the option shows Unavailable environment. Clear it, or pick another. A change made by a loop’s own session never starts a loop, and nothing shows in Run history.

Then

  • Run AI — starts a QA Wolf AI session with the instructions you write. See Run AI.
  • Run specific flows — runs the flows you pick. Select tags, flows named one by one, or both. Flows tagged later join automatically.
Run options includes failure investigation and the number of flows that run in parallel. See AI investigation for what happens after a flow fails. Run AI has no run options, except in a Smart Smoke Suite. A Smart Smoke Suite lets QA Wolf choose the flows from the change behind each deploy. Choose Run AI and start the instructions with /smart-smoke-suite. It works on a loop that starts on Pull request previews, not on a schedule, a pull request, or a QA Wolf event. Give preview deploys their own loop rather than widening the one that covers staging. A pull request rarely needs everything staging runs.

Create a loop

1
On the Loops page, select New loop.
2
In the Create a new loop dialog, describe the loop to QA Wolf, or pick a starting point: Nightly regression, Test each pull request, or Blank slate to start empty. Test each pull request sets up a Smart Smoke Suite on pull request previews.
3
Name the loop, and fill in When and Then.
4
Select Create loop.
Create loop stays disabled until the loop is complete, and says what is missing. Under the name, Add a description holds a note of up to 500 characters about what the loop is for. Empty it to clear the description. On a loop’s page, a new name, a new description, and the Enabled switch save at once. Changes to When and Then wait until you select Save changes.

Describe a loop

1
In the Create a new loop dialog, write what the loop should do under Ask QA Wolf to build the loop for you.
2
Select Ask QA Wolf. The New loop page opens with QA Wolf’s chat beside it, and QA Wolf fills in the loop’s name, When, and Then.
3
Check the loop, and change anything that isn’t right. Nothing is saved yet.
4
Select Create loop.
Example:
When QA Wolf writes a loop’s Run AI instructions, a /smart-smoke-suite command included, the loop is created paused. The page says so beside Create loop. Read the instructions, then turn the loop on with its Enabled switch. A loop that runs AI with your instructions runs its sessions with your access, because creating it makes you its owner. A loop that runs specific flows is created enabled.

Run now

Run now fires a loop by hand, so you can try it before it fires on its own. It sits on each row of the Loops page and on the loop’s own page.
  • A schedule loop — fires at once, in its environment.
  • A deploy loop — opens a menu of recent deploys the loop matches. Pick one, and the loop fires on it, in that deploy’s environment.
QA Wolf confirms that the run or the session started. The firing then shows in Run history, and a session also shows in the chat. Run now works on a paused loop, because a pause only stops a loop firing on its own. Run now is not available for a Smart Smoke Suite, a pull request loop, or a QA Wolf event loop. The button is disabled. Run now also has a cap. See Limits.

Run history

A loop’s page shows Run history, one row per firing, newest first. A row says whether the loop fired, what happened, and when. A firing started with Run now says who started it.
  • A run — opens into its results.
  • A session — shows its status, such as Session working or Session waiting for its owner. It opens into the session’s latest reply, how long it worked (“Worked for”), the flow it left if any, and Open in chat.
  • A skipped firing — says why nothing started, such as “This loop has no owner on the team to run the session as” or “This loop reached its daily limit of sessions”.
A deleted loop’s runs stay on the Runs page, and its sessions stay in the chat.

Mobile apps

A mobile pipeline uploads each new build and reports it as a deployment. The runs a deploy loop starts install that build. A scheduled run installs the environment’s latest build. When Android and iOS both report into one environment, give each platform its own loop with a service condition. See Mobile build testing.

Limits

Loops have daily and hourly caps to keep a loop from running away. A firing a cap skips shows its reason in Run history, and Run now says when a loop was run by hand too often.

Pausing

The Enabled switch on a loop’s page pauses and resumes it. Each row’s menu on the Loops page has Pause or Resume too. A paused loop never fires on its own. Resuming brings back the same configuration. A resumed schedule picks up at its next slot, and slots it missed while paused don’t run.

Use loops from a coding agent

The public API calls a loop a trigger. A coding agent connected through the QA Wolf MCP can manage loops with these operations:
  • trigger.create and trigger.update — set up a loop that starts on a deploy or a schedule. An action kind of startAgent, with instructions, makes it Run AI. Set isPaused to create it paused, so it never fires until a person turns it on.
  • trigger.runNow — fires a loop once, like Run now. A deploy loop needs the deploymentId of a recent deploy it matches. A schedule loop takes none. Send a new requestId for each firing: repeating one fires the loop once.
  • trigger.listFirings — reads a loop’s Run history, newest first: each firing’s run or session, the session’s status, and the reason for a skip.
  • trigger.find, trigger.get, trigger.pause, trigger.resume, and trigger.delete — find, read, pause, resume, and delete loops.
A loop that runs AI and is created with a workspace API key has no owner, so QA Wolf keeps it paused until a person selects Make me the owner on its page. A loop created with your own credentials is yours. The API can’t create a loop that starts on a pull request or a QA Wolf event. See Trigger fields for every field.

Reference

  • Run AI — writing instructions, and who a session runs as.
  • Trigger fields — conditions, schedule, and action fields.
  • Deployments — how QA Wolf reads a deployment, and the verdict it records for each loop.
Older workspaces may still run legacy triggers, which attach to a single environment. See Migrate from legacy triggers to move them.
Last modified on November 20, 2026