How scheduled tasks works

This guide explains what actually happens when a scheduled task runs - what the agent can access, why it never asks questions mid-run, and how results get delivered.

A scheduled task is a saved prompt that your agent runs automatically, on a repeating schedule, without anyone starting it.


Full access, same as a live chat

A scheduled run has everything your agent has in a live chat: your full knowledge base, its Actions, and its integrations. For example, it can post to Slack or update something in another connected tool, all on its own.


It completes the whole job on its own

A scheduled task runs unattended by design. It takes your saved prompt and runs it start to finish, in one shot, without pausing partway through.

📘

Note: Write your prompt as a complete instruction - say exactly what to do, where to get the information, and what to do with the result.


The prompt controls where results go

Tell the agent what to do with the result, and it does it - for example, "...then email me the summary" or "...post it to our #reports Slack channel." Without that instruction, the agent still completes the prompt and produces an answer - it just doesn't send that answer anywhere.

Every run is saved as a conversation in your agent's history, so you always have a record of what ran and when, even if nothing was sent.

🚧

Note: The agent can only deliver to a platform it's already connected to. Connect Slack, email, or wherever else you want results sent as an Action, either through Quick Setup, a custom API, or an MCP connector, before writing that instruction into your prompt.


A run keeps going once it starts

Usage limits work the same way for a scheduled run as for any message you send. Refer to How usage limits work for the details.

Turning off Plan & Act doesn't stop a run already in progress. It disables the agent's scheduled tasks going forward, and turning Plan & Act back on doesn't resume them automatically - refer to Manage your scheduled tasks to resume each one.


Check your conversation history to confirm a run worked

Your agent's conversation history shows exactly what a run did, including any that didn't produce what you expected. A scheduled run doesn't retry automatically or send an alert if something goes wrong partway through.


Availability

Scheduled Tasks is available on Premium and Enterprise plans, for agents with Plan & Act enabled.


Related articles


Did this page help you?