Routines and execution history

Work with agentsEnglish· 3 min read· Updated October 01, 2026

Repeat a tested job and follow what happened in each run.

Choose the task and how it runs

A routine repeats a task using its saved instructions and schedule. Choose its model, describe the result and configure the execution limits. You can run it without assigning an agent; MaShop does not create a hidden agent for that choice.

Assign an existing agent when the task needs that agent’s job description, selected skills or personal memory. Without an assigned agent, the routine uses its own instructions and model, with the project’s shared business rules and available tools. It does not borrow another agent’s personal memory or skills.

Test the task with Run now before relying on its schedule. State what it should read, produce and save, and which actions need your approval. A working connection, the necessary permissions and available credits are still required.

Choose the time and timezone

The scheduled time is read in the selected timezone. For example, a team working in Dakar can use Africa/Dakar for its morning report. Check that value explicitly when another teammate configures the routine from a different timezone.

A routine proposal in the chat is a configuration to review. It does not become a saved active routine merely because the assistant displayed it.

  1. Open the routine editor and name the task.
  2. Choose a one-time, daily or weekly schedule from the available choices. A weekly schedule also needs the intended days.
  3. Set the time and timezone for the people using the result.
  4. Inspect the upcoming executions shown by the editor and resolve any invalid timezone or past-date message.
  5. Review the instructions and confirm creation when you are ready for the schedule to become active.

Check a manual execution

Queued means the run is waiting to start; running means it is still working. An action awaiting approval has not finished. A completed run needs a usable output, not just a notification or an elapsed timer.

  1. Use Run now for the routine you want to test.
  2. Open its execution history and inspect the current state and recorded actions.
  3. Handle a pending approval or a missing connection if the run reports one.
  4. Check the final output in MaShop and in the destination service when the task changes external data.

Change or pause an existing routine

Open the existing routine rather than create a second copy with the same job. Pause it before changing recipients or actions that affect real data, edit its instructions or schedule, then test the result before resuming.

The assignment choice, including running without an agent, is fixed after creation in the current editor. To assign a different agent, configure a new routine and pause the old one so both do not perform the same action.

If a run fails, inspect the recorded cause and the connected service before retrying. An uncertain external result may already have happened; verify its destination before sending the action again.

Next steps

AI agents
Did this guide help?
Please leave passwords and private account details out of your feedback.