---
title: "Agent Tasks: schedules and triggers"
description: Understand timezone-based schedules, activation, manual runs, events and overlapping work.
canonical: https://www.proxyai.app/docs/agent-tasks-schedules
last-updated: 2026-10-19
---

> For the complete documentation index, see [llms.txt](https://www.proxyai.app/llms.txt). Request public content pages with `Accept: text/markdown`.

# Agent Tasks: schedules and triggers

## Scheduled runs

An active scheduled task has a next-run time calculated from its schedule and timezone. The scheduler periodically checks due tasks and queues the active compiled version. The displayed time is a scheduled target, not a guarantee that execution starts at that exact second.

Schedules use the selected timezone. Daylight-saving changes can affect the matching local time; a local time skipped by a spring-forward change does not occur. Inspect the next-run time after changing a schedule or timezone.

Library tasks become active when added. A deployed custom task uses the active version. Pausing stops automatic scheduling; resuming restarts it.

## Overlapping work

The scheduler checks for an open run before queueing another scheduled execution. Queued, claimed, running, held and applying runs count as open. A held approval can therefore prevent a later scheduled run from starting.

The scheduler advances to a future occurrence rather than creating a separate catch-up job for every missed time. The runtime also checks concurrency when claiming work. Inspect the Run Log and any waiting proposal if a scheduled task appears delayed.

## Manual runs

Run now queues one execution and returns a run identifier. A paused task with an executable version can still be run manually; pausing its schedule does not prohibit an explicit manual run. Archived tasks cannot be run. A request can be refused when another run is already open.

A live manual run uses the active version. A test uses the latest compiled version. See [test-mode effects](https://www.proxyai.app/docs/agent-tasks-execution) before starting either.

## Events and webhooks

Event-triggered tasks respond to the event configured in their trigger, such as a relevant ticket, chat or supported store event. Available events depend on the connected platform. A webhook-triggered task receives its own address in the console; use the displayed setup instructions rather than reusing a URL from another task.

## Customer requests in chat

A deployed chat-callable task advertises its name, when-to-use instruction and input contract to the bot. The customer request starts a run subject to its identity, allowance and tool restrictions. Pausing unpublishes the task from chat.

See [permissions and customer data](https://www.proxyai.app/docs/agent-tasks-permissions) before exposing any task to shoppers.