> ## Documentation Index
> Fetch the complete documentation index at: https://docs.longoz.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Routines

> Schedule repeatable work, run it on demand, and follow each generated Issue.

A Routine creates a new [Issue](https://docs.longoz.ai/issues.md) from a saved template and assigns it to an Agent. Use it for work you want repeated, such as a weekly operations summary or a daily check of local reports.

Routines support a schedule and **Run now**. There is no webhook trigger in the current UI.

## Before you create one

You need a ready [Project](https://docs.longoz.ai/projects.md) with a healthy Agent assigned to it. The Agent must use **Full** autonomy. Its [Computer](https://docs.longoz.ai/computers.md), provider authentication, and required tools must be ready to execute the task.

Full autonomy permits future runs to execute without approval prompts. Choose an appropriate Computer and give its tools only the access the task needs. A reviewer checks the result after execution; assigning one does not turn Full autonomy into approval before each action.

Owners, admins, and members can create, edit, and run Routines for Projects they can access. Viewers can read them.

## Create a weekly operations summary

This example asks your configured Agent CLI to read CSV exports already present on its Computer. Longoz does not provide a native connection to your ERP or database. If your task uses an external service, install and authenticate the necessary tool on that Computer first.

1. Use a Project with a repository containing `reports/orders.csv` and `reports/support.csv`. Check that the exports contain dates and the fields you want summarized.

2. Open **Create issue**, select the Project, and assign its Full-autonomy Agent.

3. Enter `Weekly operations summary` as the title. Add a description such as:

   ```text theme={null}
   Read reports/orders.csv and reports/support.csv in this Project.
   Summarize the previous Monday through Sunday using Europe/London dates.
   Report the number of orders, total order value, and open support items
   if the source columns support those calculations. State the date range
   and source filenames. If a file or required column is missing, report
   what is missing instead of estimating values.
   Write the summary to weekly-operations-summary.md in the Project and
   include the findings in your final response. Do not change the input files
   or send anything to an external service.
   ```

4. Enable **Routine**. Choose **Recurring**, then **Weekly**. Select Monday and set **Recurring time** to `09:00`.

5. Set **Timezone** to `Europe/London`. Review **Next 5** to check the intended dates and times.

6. Open **Advanced**. Keep **Do not overlap** and **Skip missed runs** for a report that should not build a backlog.

7. Choose a reviewer if someone should check each report. Without a reviewer, a successful occurrence moves its Issue to **Done** automatically, including integration of any code changes. With a reviewer, success moves it to **In review**.

8. Select **Create routine**.

The saved description is the Issue template. Each occurrence gets a separate Issue with the Routine's title, description, Agent, priority, labels, attachments, and reviewer. Generated Issues start in **Todo** with Standard work mode. Routines do not carry acceptance criteria or create their occurrences under a parent Issue.

Update and commit the exports before each scheduled run, or arrange that separately with tools you already operate. Scheduling the summary does not refresh its source files.

## Run it once and inspect the result

1. Open **Routines** and select `Weekly operations summary`.
2. Select **Run now**. This creates a real occurrence, so use inputs and access that are safe for the Agent to use.
3. Find the occurrence under **Runs** and open its linked Issue.
4. Review the Agent's response. A report file saved in its workspace is not automatically an Artifact; ask the Agent to publish it if you need a downloadable output. If you assigned a reviewer, complete the review before the next occurrence when using **Do not overlap**.

**Runs** shows the source, scheduled time, actual start, result, linked Issue, and revision. Use it to distinguish an Issue that was created from work that has actually succeeded. A skipped or configuration-blocked occurrence may have no new Issue to open.

**Run now** is available for active and paused Routines. It still checks the Agent and Project configuration and obeys the overlap policy. Archived and completed Routines cannot run again.

## Choose a schedule

Choose **Once** for a future local date and time. Choose **Recurring** for hourly, daily, weekday, weekly, or monthly work. **Custom cron** accepts a standard five-field expression with numbers, ranges, lists, wildcards, and steps.

For example, `0 9 * * 1-5` means 09:00 Monday through Friday in the Routine's selected timezone. The schedule preview is the place to check the resulting occurrences before saving.

Use an IANA timezone such as `Europe/London` or `America/New_York`, not a fixed UTC offset. Your personal date and time preferences change how timestamps are displayed; they do not change the Routine's schedule.

## Handle overlapping and missed work

Under **Advanced**, choose what should happen when earlier work remains open:

* **Do not overlap** skips an occurrence while a previous linked Issue is in Backlog, Todo, In progress, In review, or Blocked. Waiting for a review counts as active work.
* **Queue every occurrence** creates work for each occurrence even when an earlier Issue is still open. Agent capacity still controls when the work can execute.

The missed-run setting controls occurrences that the scheduler processes late:

* **Skip missed runs** leaves overdue occurrences unexecuted.
* **Run the latest missed occurrence once** creates one catch-up occurrence instead of replaying every missed slot.

For a report, skipping missed work often avoids producing stale summaries. For a periodic inspection, one catch-up run may be useful. Choose based on what the task should do with old inputs.

## Edit, pause, or stop a Routine

Open the Routine and select **Edit** to change its template, assignment, reviewer, attachments, or schedule. **Agent details** holds execution settings and [Skills](https://docs.longoz.ai/skills.md). Changes create a revision for future occurrences; existing Issues and their history remain unchanged.

Use **Revision history** to inspect earlier revisions. **Restore** creates a new head revision from the selected snapshot. It does not rewrite existing revisions or Issues. Archived and completed Routines cannot be edited or restored.

Select **Pause** to stop future scheduled claims. Existing Issues remain available. Select **Resume** when the configuration is ready again; recurring schedules continue from the next future occurrence, without replaying the paused period. A one-time schedule whose date has passed needs a future date before it can resume.

Pausing the Agent, or reducing its autonomy from Full, also pauses its active Routines. Reactivating the Agent does not resume them automatically. Check each Routine and resume it explicitly.

**Archive** stops future scheduling and preserves existing Issues and history.

## Another task to schedule

For a daily data-quality check, ask your Agent CLI to inspect a CSV export in the Project, list rows with missing identifiers, and write a report without modifying the source. Select **Every day**, choose the business timezone, and run it once manually before leaving the schedule active. The export and any parsing tools must already be available on the assigned Computer.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.