Skip to content
CleroHelp center

Search documentation

Try a topic or a task

↑ ↓ Move through results · Enter Open · Esc Close

Русский
English · Review draft
Documentation previewReview draft · English

Schedule recurring work

Ask an agent to create, inspect, change, and pause a recurring schedule.

Tested in part Evidence & limitations

Live agent-created inactive recurring schedule and follow-up read verified; execution, timing outcomes, and lifecycle changes source-backed. This guide is a review draft; check access and saved results in your current workspace.

On this page

Ask the responsible agent when you want work to repeat. Describe the task, timetable, time zone, and result destination; then check the saved schedule. You do not have to begin with an administrator's automation editor.

Use a scheduled task for one action at one future time, a trigger for defined work from eligible input, or a recurring schedule (cron job) for a repeated task. For a general project-manager review, use Heartbeats. For manager checks caused by objective state changes, use Project Flow.

Choose the right place to ask

For project work, open the project and ask the agent responsible for the work. The project, agent, sources, and tools must all be eligible for that workflow.

For a personal reminder, use Space assistant. Home schedules run as that assistant in your Home conversation. They cannot be reassigned to a project or another agent, or redirected to an external report destination. To change a project workflow from Home, ask the assistant to coordinate with the project manager instead.

Create a recurring schedule through chat

  1. Open the correct project conversation or Home.
  2. Describe the repeated action and name its authoritative sources.
  3. State the frequency and an explicit time zone, such as Europe/Berlin or UTC.
  4. Specify where the result should appear and which external actions are allowed.
  5. Ask for a saved confirmation containing the name, assigned agent, timing, time zone, active state, and destination.
  6. Ask the agent to read or list the saved schedule again before relying on it.

For a reviewable setup before activation:

Create a recurring schedule called Weekly project check for Mondays at 09:00 UTC. Read this project's checklist and summarize unfinished items. Keep it paused from the first write, with no external report destinations. Do not run it now. Confirm the saved schedule and that Active is off.

Do not assume every creation path defaults to paused. The current source supports an explicit inactive state, but Home creation can default to active. If the agent cannot guarantee the requested paused setup, have it explain the blocker rather than creating an active substitute.

To start the intended recurrence after review:

Activate Weekly project check using the saved instructions and Monday 09:00 UTC schedule. Confirm its saved active state and timing.

The documentation demo completed this paused-setup pattern after explicit capability approval. Documentation Lead created Documentation Demo — Disabled Weekly Check for Mondays at 09:00 UTC. The create result and a separate read both showed active: false; validation passed, and the record had never run. No schedule was activated and no report was delivered.

Although the request omitted report destinations, the saved record attached the existing manager conversation automatically, matching Persistent team behavior. The example name above is adaptable; confirm the exact name and state in your own saved record.

Clock time, interval, and cron

Clock recurrence means a named local time, such as weekdays at 09:00 Europe/Berlin. Interval means elapsed time, such as every two hours. Every 24 hours is not necessarily the same instruction as every day at 09:00 local time.

A cron timetable uses minute, hour, day of month, month, and day of week fields. For example, 0 9 * * 1 describes Mondays at 09:00 in the schedule's configured time zone. You can give the timetable in ordinary language; the agent can create the appropriate fields.

Current source supports new CRON and INTERVAL schedules and enforces at least 15 minutes between recurring runs. New one-off actions use the separate scheduled-task mechanism; older one-off cron records may still exist.

For several specific times, state complete pairs: “09:30 and 16:45,” rather than separate lists of hours and minutes that could produce extra combinations. Named time zones are supported; daylight-saving skipped/repeated times, catch-up, and exact execution timing have not been tested in this guide. Check the saved timetable and representative runs before relying on critical timing.

Know where results go

  • Persistent team project: The assigned agent's ongoing project conversation is attached automatically. Omitting destinations does not remove that conversation.
  • Flexible chats project: A compatible agent/project chat can be connected. Without one, independent execution can remain in run history.
  • Home: Execution and result stay in that original Home conversation.
  • External reports: Only configured, supported destinations are used. A connected service is not automatically a report destination.

The final schedule response is delivered by the reporting mechanism. Avoid separately instructing the agent to post the same report again, which can duplicate delivery. A saved schedule, a successful run, and successful report delivery are three different checks.

Inspect and change the existing schedule

Ask by its name or saved ID:

Show Weekly project check, including paused state, instructions, agent, time zone, and result destination.

Change Weekly project check to Fridays at 16:00 Europe/London. Keep its instructions and destination unchanged, and confirm the saved result.

The current tool route is create, list/read, update, and delete through create_automation_cron_job, list_automation_cron_jobs, read_automation_cron_job, update_automation_cron_job, and delete_automation_cron_job. Users do not need to type these tool names; they explain why asking an agent is a real configuration path rather than a reminder written only in chat.

Pause future runs or stop current work

Pause Weekly project check so it will not start future scheduled runs. Tell me whether any execution is already active.

Pausing changes the definition's active state and disables future periodic starts. It does not cancel an execution already running. If you intend both, explicitly request the pause and use the active run's Stop control, when available, to stop current execution.

Deleting the schedule removes the definition; it is different from pausing it for later reuse. Deleting it also does not stop an execution already running. Neither pause nor delete reverses completed file edits, messages, or other external effects.

Find the workflow views

Open project Workflows → Schedules for recurring definitions, or Scheduled tasks for one-off activity. The workspace Workflows view can show work across the current context. Triggers is the event-workflow view. Heartbeats contains adaptive Idle check and one-time/daily project-manager checks; those are user-managed checks with different lifecycle controls, not ordinary agent-created cron jobs.

In the live Member demo, the Schedules view still showed No schedules yet after the agent created and read back the disabled record. The linked artifact also did not reveal a usable editor. This visibility discrepancy is unresolved: ask the agent to read the known schedule and report its saved state rather than assuming the empty list means it does not exist.

Source-backed list controls include Activate/Deactivate, Edit, and Delete when permitted. Home resources have their own personal-schedule controls. The editor can use legacy wording such as Create Scheduled Task under Schedules; choose the mechanism by the intended behavior, not that label alone.

If an editor says No Access

The live Member account could not open the new-trigger administration form. That is a restriction on that route, not proof that Members cannot ask agents to create their own schedules or triggers. Current backend permission paths support Member-owned definitions with the required space/project access and agent tools.

If the agent reports a missing capability or denied approval, resolve that exact blocker. Asking through chat does not bypass authorization. Do not switch identities or silently expand scope. See Capabilities, Roles and access, and Plans and usage.