Routines: schedule AI tasks that run on your machine
A routine is a saved prompt on a schedule — daily, weekly, hourly or on demand. It runs headlessly through your own key, saves the result as a conversation and notifies you when it’s done.
The More AI team · Updated June 14, 2026
What a routine is
Think of a routine as a standing order for your assistant: “every morning, summarize what changed in this folder”, “every Friday, draft a weekly status from my notes”, “each hour, check this repo’s TODOs”. You write the prompt once, pick a schedule, and More AI runs it in the background — locally, through the same engine and key as a normal conversation.
Each run is saved as a full conversation in an inbox-style history, with unread badges, so you can review what the agent did run by run.
Creating your first routine
Routines live in two places: Cowork → Scheduled (for project tasks) and Code → Routines (for repository tasks). Click “New routine”, give it a name and a task prompt, pick the project or repo folder it may read, choose the model and the frequency — interval, daily, weekly, once, or manual.
A keep-awake toggle uses a real power-save blocker, so scheduled runs actually fire on a laptop that would otherwise sleep.
What routines may (and may not) do
Because no one is watching a headless run, routines deliberately run with read tools only — gated, destructive actions are withheld. With web search enabled, a routine can also pull pages from the internet, which makes patterns like a “morning digest” work with zero extra services.
Ideas to steal
A few patterns that work well in practice:
- Morning digest: search the web for news on your topics and summarize the top items.
- Weekly review: read this week’s notes in a project folder and draft a status update.
- Repo watch: scan a repository for new TODO/FIXME comments and list them by file.
- Content pipeline: turn yesterday’s raw notes into a cleaned-up outline every night.
Frequently asked questions
What can a routine do on its own?
Routines run headlessly with read-only tools — destructive, gated actions are withheld because no one is watching. With web search enabled they can also pull pages from the internet.
Will routines run if my laptop is asleep?
A keep-awake toggle uses a real power-save blocker, so scheduled runs actually fire even on a laptop that would otherwise sleep.
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?
Anything you’d otherwise do on a schedule from your files or the web — a morning news digest, a weekly status drafted from your notes, an hourly scan of a repo’s TODOs, or turning raw notes into a clean outline each night.
Do routines use my API key and cost money?
Yes — a routine runs through the same provider and key as a normal conversation, so each run costs the tokens it uses. With a local model there is no per-run cost.