Scheduled Prompts

A scheduled prompt is a message Loca sends on its own, on a schedule you set, so the agent works for you when you are not watching. It can run once at a future time, every so often, or on a weekly, monthly, or yearly cadence. Each run opens a fresh session, so the result is waiting for you as a conversation you can read and continue, not a notification you have to act on in the moment.

A scheduled prompt is an ordinary prompt, so anything you can ask in a session you can ask on a schedule: a morning digest, a weekly summary, a recurring check against your memories. The fresh session per run keeps each result clean and self contained.

Setting one up

Add and edit scheduled prompts under Scheduled Prompts in Settings, where you write the prompt and choose its cadence.

The agent can also propose a schedule mid-conversation. Ask it to remind you of something or to check back later, and it can offer to set up the scheduled prompt for you, which you approve right there in the session. That way a recurring task can come out of a normal conversation instead of a trip to Settings.

Missed runs

A schedule is only dependable if it survives downtime. If your Core was off when a prompt was due, Loca can make up the missed run the next time it starts, rather than silently skipping it. So a daily digest you rely on still arrives after your machine was asleep overnight, instead of leaving a gap you might not notice.

How It Works

A scheduled prompt is the one Loca feature that runs with no app open, so it lives entirely on the Core. The Core keeps each schedule with its cadence and works out when the next run is due; when that time arrives, it starts a session by itself, runs your prompt as a turn using its own default settings, and saves the result for you to find later. There is no Client in the loop, which is why a scheduled prompt keeps working while your laptop is closed, as long as the Core is running. The make-up behavior falls out of the same bookkeeping: on startup the Core compares the current time against when each prompt was last due, and runs any that should have fired while it was off.

That is the last of the automation features. The next chapter is the tools that let the agent act, starting with Tools.