> ## Documentation Index
> Fetch the complete documentation index at: https://docs.conduit.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Schedules

> Run an internal agent on its own at set times, such as every morning or every few hours, and manage the follow-ups an agent schedules for itself.

A schedule runs an internal agent at times you set, with a task you write once. Use schedules for recurring jobs such as a morning arrivals check, an hourly sweep for unanswered messages, or a weekly summary for your team.

Open the agent and select **Runs on > Schedule**. The sidebar row shows how many schedules are active, or **Setup** when none are.

> [!NOTE]
>
> To run an agent a set time before or after each reservation's check-in or check-out, use a [Check-in or Check-out platform event](https://docs.conduit.ai/build-your-agent/agents/internal-agents/platform-events/reservations-and-appointments.md#check-in-and-check-out) instead of a schedule.

## Create a schedule

![The New schedule page with Every day selected, a 9:00 AM run time, the America/Los\_Angeles time zone, and a preview of the next three runs](https://mintcdn.com/conduit-58d85182/Dxa79236R9TzscJw/images/internal-agents/new-schedule.png?fit=max\&auto=format\&n=Dxa79236R9TzscJw\&q=85\&s=58f148e731f64952c615295fa577b9da)

**1. Add a schedule**

Select **Add schedule**.

**2. Choose when it runs**

Under **When should it run?**, pick **How often**, then fill in the times. See [Schedule options](#schedule-options).

**3. Pick the time zone**

Choose the **Time zone** the schedule follows. The preview below it shows the cadence in words and the next few runs under **Coming up**.

**4. Tell the agent what to do**

Under **What should the agent do?**, write the task. The agent gets these instructions every time this schedule runs, so write them the way you'd brief a teammate. To start from an example, select **Arrivals check**, **Unanswered messages** or **Weekly summary**.

**5. Try it**

Select **Try it now** to run the agent once with this task and see what it does. Follow the run in the **Activity** tab.

**6. Save**

Select **Create schedule**.

## Schedule options

| How often | Example | What you set |
| - | - | - |
| **Every day** | Every morning at 8 | One or more times of day. |
| **On certain days** | Weekdays at 9 AM | **Which days**, and one or more times. |
| **Every few hours** | Every 2 hours, 8 AM to 8 PM | The hours between runs, and the first and last run of the day. |
| **Once a month** | On the 1st at 9 AM | **Which day of the month**, and one or more times. |
| **Just once** | On a date you pick | **Which date** and a time. |

For daily, weekly and monthly schedules, select **Add another time** to run more than once a day, up to 6 times.

**Every few hours** runs at most every 12 hours, and only between the first and last run you set. This keeps the agent from running overnight while guests and your team are asleep.

For **Once a month**, months without the day you picked are skipped. Pick **last day** to always run at the end of the month.

A **Just once** schedule runs one time, then turns itself off.

## Write a good scheduled task

A scheduled run has no message to react to, so the task is all the agent knows about why it's running. Make it specific:

* **Say what to look at.** "Today's arrivals", "conversations waiting more than an hour", "tasks due tomorrow".
* **Say what done looks like.** "Post one Slack message listing each listing and its status."
* **Say what to skip.** "Ignore owner stays", "Don't message guests directly".

Keep long procedures in a [Skill](https://docs.conduit.ai/build-your-agent/skills.md) and refer to it from the task.

> [!TIP]
>
> A scheduled agent can only make so many tool calls in one run. If each run has to read a lot of conversations, run it more often over a shorter window, or have your customer agent tag the conversations it should look at. See [Working within a run's tool budget](https://docs.conduit.ai/build-your-agent/agents/internal-agents/use-cases.md#working-within-a-runs-tool-budget).

## Manage schedules

Each schedule in the list shows its cadence, its task, and a status line. Select a schedule to edit it.

| Action | How |
| - | - |
| **Pause or resume** | Use the switch on the schedule. |
| **Run now** | Open the schedule's actions menu and select **Run now**. The run appears in **Activity**. |
| **Delete** | Open the schedule's actions menu and delete it. The agent stops running on that schedule. Past runs stay in **Activity**. |

An agent can have up to 20 active schedules.

### Schedule status

| Status | Meaning |
| - | - |
| **Next run** | The schedule is active and shows when it runs next. |
| **Paused** | You paused the schedule. |
| **Paused after repeated failed runs** | The schedule failed 3 runs in a row, so Conduit paused it. Check the failed runs in **Activity**, fix the cause, then resume it. |
| **Ran once** | A **Just once** schedule that has run. |
| **Ran once and ended with an issue** | A **Just once** schedule whose run ended with a problem. |

Resuming a schedule clears its failure count and works out the next run from now.

## Follow-ups set by the agent

An agent can also schedule follow-ups for itself during a conversation, for example "Check again tomorrow at 10 whether the cleaner confirmed." These appear under **Set by the agent** on the **Schedule** page.

When a follow-up fires, the agent reopens that conversation with the note it left and decides what to do next, such as replying or checking again. Each follow-up shows its note, which conversation it belongs to, and when it was set. To stop one, open its actions menu and cancel it.

Follow-ups use Conduit's built-in scheduling tool. See [Built-in tools](https://docs.conduit.ai/build-your-agent/tools.md#built-in-tools).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.