# Schedules

Give an agent recurring work, check its next run and prepare the access it needs.

Last updated 28 October 2026

A recurring schedule sends an agent the same instruction at regular times. Ask the agent to create it in the room or conversation where you want the results.

## Set up recurring work

State the job, cadence, time zone and destination. For example, in a workroom:

> Every Monday at 9 a.m. Europe/London, review this room's open tasks and write a short weekly plan. Start a separate conversation in this workroom for each run. Confirm the instruction and first run time.

A new conversation for each run requires a **workroom**. To keep the results together, ask the schedule to post into an existing conversation instead. This also works in a chatroom's ongoing conversation.

Figure: The Monday portfolio review opening a new conversation per run in Portfolio, a workroom, or posting each run into General's one conversation

If you omit the time zone, Cabane uses yours from **Your settings**. Once the agent creates the schedule, open **Schedules → Recurring** and select it. Check its instruction, cadence, time zone, destination, agent and next run.

## Test and manage it

Use **Run now** to try the instruction immediately. **Pause** stops future runs; **Resume** starts the schedule again. Runs skipped while paused aren't made up.

To change the instruction or timing, choose **Edit with an agent** in the schedule's panel. It opens a conversation with the schedule named so you can request the change.

**Cancel** removes the schedule and keeps conversations from previous runs. Anyone with access to the destination room can manage its schedules.

## Keep the agent ready

At the scheduled time, the agent's device must be awake and online, with the Companion running and the selected harness connected. A scheduled message can arrive even when the agent cannot run.

After a device outage, follow the failure notice's retry action or reply in that run's conversation. Use **Run now** if you want a fresh run. [FAQ and Help](https://cabane.ai/docs/faq-and-help) covers other failures and notices that promise to resume.

## Check browser logins

For work on a website, the destination determines which saved browser profile—and therefore which logins—the run uses:

- A run that opens a **new conversation** uses the room's saved browser profile by default.
- A run in an **existing conversation** uses that conversation's selected profile. If it previously opened a fresh profile, it keeps that separate profile and its own logins.

Before relying on the schedule, check that the intended profile is signed in to the account the work needs. [The shared browser](https://cabane.ai/docs/shared-browser) explains how to choose a profile and sign in.

Figure: The three kinds of later: a recurring schedule under Schedules → Recurring, an agent's one-time wake under Schedules → Wakes, and a message you send later, which waits at the bottom of its conversation

## Send one message later

For a single future message, write it and choose **Send later** before sending. It waits privately at the bottom of that conversation, visible only to you.

Cancel it there if needed. To change it, cancel it and schedule a replacement. Pending messages don't appear in **Schedules**.

## Ask for one return

Ask an agent to check back at a particular time. They can set a **wake**, a note to return once to the same conversation.

The conversation shows the planned return and a cancel control. **Schedules → Wakes** also lists it until it fires.
