Scheduled AI tasks: local routines and cloud check-ins
Create one-off or recurring AI tasks, choose local or cloud execution, and review their results. Practical ideas for research, page checks and weekly reports, with clear schedule and file-access limits.
The More AI team · Updated September 14, 2026
What a routine is
A routine is a saved instruction that runs again on a schedule. A one-off scheduled task runs once. Use either for a concrete result: a weekly reading list with links, a report from selected notes, or a check of specified public pages.
Choose where it runs. A local routine uses the desktop app and needs that computer available. A cloud routine runs on More AI infrastructure and can work while your laptop is closed. Cloud work requires an account and a model available for that account. Each run leaves a conversation you can inspect.
Creating your first routine
Open Scheduled and describe what should happen and when, or choose a suggested task. The assistant can ask about missing details and prepare a cloud schedule. Manual setup remains available; repository-bound local tasks have their own routine controls in Code.
Before saving, check the execution location, sources, time zone, model and expected output. Run the task manually first. A keep-awake option helps local routines avoid ordinary sleep; it does not turn a powered-off computer into a cloud runner.
What routines may (and may not) do
A routine uses the permissions and execution mode you configure. Read-only or ask modes withhold actions that need interactive approval; configured unattended or Auto/Full access can permit writes and commands within their remaining restrictions. Start with a manual read-only run and inspect its available tools. Scheduling does not make a routine harmless or guarantee that every action will complete.
Useful tasks to schedule
Begin with one task whose result you can check in a few minutes. The app includes ideas for weekend reading, discounts, nearby concerts and weekend plans; adapt the same pattern to your work.
- Weekly industry digest: choose topics and sources; ask for five links, dates and what changed.
- Public-page review: check named supplier or competitor pages and record observed prices or terms, with a timestamp and source links.
- Report from prepared files: select or upload the current notes or spreadsheet, then request a summary and exceptions to review. Decide who will update the source before the next run.
- Local repository review: list new TODOs or summarize changes in a selected repository while its computer is available.
- Personal shortlist: specify a country, city, interests and budget for discounts, concerts or weekend ideas.
A read-only public-source review to try manually
The downloadable kit checks three public More AI pages: guides, documentation and release security reports. Its prompt asks for a timestamp, exact source URLs and a concise summary, with no sending, publishing or file changes. The first run records a baseline; it must not invent changes without a previous snapshot.
The kit contains a prompt and a checklist, not a run log. No scheduled or cloud run was executed to produce it. Security findings must come from the current report; a missing report means unavailable, not a clean verdict.
- Open the prompt and review the allowed public sources. Enable only the read and web tools this task needs.
- Run it manually first. Open its source links and confirm the timestamp, findings and unavailable-page handling.
- Save a verified baseline. On later runs, compare only when that baseline is accessible to the execution context; otherwise report that comparison is unavailable.
Verify the execution context before scheduling
A local routine needs its computer available. A cloud routine requires an account, an available model or usable cloud provider key, and context the cloud can access. A desktop path does not automatically exist on the server. Model tokens and any search charges follow the selected services; cloud beta availability is temporary.
- Confirm local or server execution, model, allowed tools, timezone and next-run time.
- For this read-only recipe, disable writes and unrelated connectors rather than relying only on prompt wording.
- Check provider quota and errors; an empty result is not proof that nothing changed.
- Inspect the first scheduled run and record failures. Pause or remove the routine when its sources or purpose change.
Schedules are periodic checks, not exact-time alarms
Cloud schedules are checked every 15 minutes. A run can start after the selected time, and the result also takes time to generate; queueing and service availability can add delay. Use these tasks for periodic research and reports, not time-critical alarms.
Keep the intended time zone explicit. After the first scheduled run, open its result and check the sources, timestamp and any errors. An empty answer or an inaccessible page does not prove that nothing changed.
Give the cloud the context it can actually use
A desktop folder, local model, installed command or signed-in desktop browser does not automatically become available to a cloud routine. Select or upload the required files and use connectors supported in the cloud. A comparison with an earlier run needs an accessible baseline; otherwise ask for a fresh snapshot.
Add a provider key for cloud use separately, or choose a model made available to your account. Available cloud features are temporarily free during beta; model usage through your own key is still billed by that provider. You can pause a routine, change its schedule or remove it when it is no longer useful.
Frequently asked questions
What can a routine do on its own?
A routine uses the permissions and execution mode you configure. Read-only or ask modes withhold actions that need interactive approval; configured unattended or Auto/Full access can permit writes and commands within their remaining restrictions. Start with a manual read-only run and inspect its available tools. Scheduling does not make a routine harmless or guarantee that every action will complete.
Will routines run if my laptop is asleep?
Choose where it runs. A local routine uses the desktop app and needs that computer available. A cloud routine runs on More AI infrastructure and can work while your laptop is closed. Cloud work requires an account and a model available for that account. Each run leaves a conversation you can inspect.
Where do I see what a routine did?
Each run is saved as a full conversation in an inbox-style history with unread badges, so you can review exactly what the agent did, run by run.
What kinds of tasks suit a routine?
Weekly industry digest: choose topics and sources; ask for five links, dates and what changed. Public-page review: check named supplier or competitor pages and record observed prices or terms, with a timestamp and source links. Report from prepared files: select or upload the current notes or spreadsheet, then request a summary and exceptions to review. Decide who will update the source before the next run. Local repository review: list new TODOs or summarize changes in a selected repository while its computer is available. Personal shortlist: specify a country, city, interests and budget for discounts, concerts or weekend ideas.
How are routines priced?
Local routines use your chosen provider key, so you pay that provider for model tokens; a local model has no token charge. Server execution belongs to the cloud subscription layer, while model tokens remain on your provider key. Cloud features are temporarily free during the current beta, with advance notice promised before billing begins.
Will a cloud routine deliver at the exact minute?
No exact-minute delivery is promised. Cloud schedules are checked every 15 minutes, then the task needs time to run. Use them for periodic checks and summaries rather than time-critical alarms.
Can a cloud routine read a folder on my closed laptop?
Only files made available to the cloud can be used there. A local folder, local model or desktop-only connector does not move to the server when you save a schedule.