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

# Tasks (Cron & Heartbeat)

> Scheduled cron jobs and heartbeat-based background tasks created by your agent

# Tasks

Aivah's **Tasks** page (`Side nav → Integrations → Tasks`, or `/integrations/tasks`) is where every scheduled or recurring job created by you or your AI agent lives. Anything from *"remind me on Monday at 9 AM"* to *"ping me if the API is down"* shows up here.

The page is split into two tabs, each with badge counters and a **Refresh** button.

## Cron Jobs

Scheduled jobs that run on a cron expression.

Each row shows:

* The cron expression and the human-readable next-run time
* The action the job will take
* The **channel** the job runs on (Web, Slack, WhatsApp, etc.)
* Inline **Run now** and **Delete** controls

Use Cron Jobs for any recurring action you want the agent to take on a schedule – sending a weekly summary, opening a daily standup thread, refreshing data sources, posting reminders, etc.

<Frame caption="Cron Jobs tab – a reminder scheduled at `0 9 * * *` on the Web channel">
  <img src="https://mintcdn.com/opinionai/tR4nAAll3jemMkvA/images/tasks-cron-jobs.png?fit=max&auto=format&n=tR4nAAll3jemMkvA&q=85&s=0e9f3b42a65d6d772559545e99cb879c" alt="Cron jobs" width="4112" height="2658" data-path="images/tasks-cron-jobs.png" />
</Frame>

### Manage cron jobs

* **Run now** – executes the job immediately without changing its schedule. Useful for verifying that the action works.
* **Delete** – removes the job permanently. The agent will no longer trigger it.
* **Refresh** – re-fetches the list so freshly created jobs appear.

## Heartbeat

Background tasks that need a regular heartbeat to remain healthy. If the heartbeat stops coming in, the task is flagged so you can intervene.

The panel shows:

* A **Heartbeat Monitor** card with the interval, channel, last beat, and next beat
* A list of every heartbeat task with its priority, status, and creation timestamp

Heartbeats are ideal for long-running monitors and watch-style integrations – think uptime checks, queue depth monitors, or external service health pings.

<Frame caption="Heartbeat tab in its empty state, before any tasks have been wired up">
  <img src="https://mintcdn.com/opinionai/tR4nAAll3jemMkvA/images/tasks-heartbeat-empty.png?fit=max&auto=format&n=tR4nAAll3jemMkvA&q=85&s=6840b813d4a0fb802ccc62c41488370d" alt="Heartbeat empty state" width="4112" height="2658" data-path="images/tasks-heartbeat-empty.png" />
</Frame>

<Frame caption="Heartbeat tab with an active monitor and one running task">
  <img src="https://mintcdn.com/opinionai/tR4nAAll3jemMkvA/images/tasks-heartbeat-populated.png?fit=max&auto=format&n=tR4nAAll3jemMkvA&q=85&s=738538f5e888f374a98019bc85f5f75d" alt="Heartbeat populated" width="4112" height="2658" data-path="images/tasks-heartbeat-populated.png" />
</Frame>

## How tasks are created

Tasks generally appear here in one of three ways:

1. **You ask the agent** – natural language requests like *"remind me to send the weekly report every Friday at 9 AM"* create a cron job automatically.
2. **An MCP tool returns a schedule** – some MCP integrations (Calendar, Linear, etc.) can schedule follow-ups that surface as cron jobs.
3. **You configure a heartbeat-style monitor** – longer-running integrations are wired up via the heartbeat configuration card.

## Workflow

<Steps>
  <Step title="Open Tasks">
    Navigate to **Integrations → Tasks** and pick the tab you want.
  </Step>

  <Step title="Inspect status">
    Use the badge counters to see how many jobs exist. Check timestamps to spot anything misbehaving.
  </Step>

  <Step title="Run a quick verification">
    For cron jobs, hit **Run now** to confirm the action still works after a code or content change.
  </Step>

  <Step title="Clean up">
    Use **Delete** on jobs you no longer need, then **Refresh** to confirm they're gone.
  </Step>
</Steps>

## Troubleshooting

* **Job didn't run on time** – use **Run now** to confirm the action itself still works, then double-check the cron expression and the agent's MCP tool access.
* **Heartbeat stopped** – inspect the last heartbeat timestamp, then check the upstream service that's supposed to be sending the heartbeat.
* **List looks stale** – hit **Refresh** at the top of the tab to re-fetch.

## Related guides

* [MCP Tools](/integration/mcp-tools)
* [Workflows](/integration/workflows)
* [Memory](/integration/memory)