Skip to main content

Overview

Zeus’s automated tasks allow you to schedule the AI Agent to execute prompts at specified times — whether it’s a daily report, weekly code review, or a one-time data processing task. Once a task completes, Zeus notifies you of the results via email, Feishu, or in-app notification.

Creating a Task

Navigate to Settings > Automated Tasks to create a new scheduled task. Each task requires the following configuration:

Basic Settings

SettingDescription
TitleA descriptive name for the task
PromptThe instruction the Agent will execute — just like sending a message in chat
ConnectorOptional — bind a Browser or Desktop connector for tasks that require UI interaction

Schedule Settings

SettingOptionsDescription
Scheduled DateDate pickerThe date for the first (or only) execution
Scheduled TimeTime picker (e.g. “09:00”)The time of day to execute
TimezoneUser’s timezoneEnsures the task runs at the correct local time
Repeat TypeNo repeat / Daily / Weekly / MonthlyHow often the task should recur

Run Mode

ModeBehavior
Auto-proceedThe Agent executes autonomously — no user intervention needed
Require confirmationThe Agent pauses before sensitive operations and waits for your approval (HITL)

Notification Settings

Configure how you want to be notified when a task completes:
ChannelDescription
EmailReceive execution results via email
FeishuReceive a message through Feishu integration
In-AppView results in the notification center within Zeus
You can enable multiple notification channels simultaneously.

Task Lifecycle

A scheduled task goes through the following states:
StatusDescription
PendingTask is scheduled and waiting for the next execution time
RunningThe Agent is currently executing the task
CompletedExecution finished successfully — results are available
FailedExecution encountered an error — error details are available
PausedTask is temporarily suspended — will not execute until resumed
CancelledTask has been permanently cancelled
For recurring tasks, the status automatically returns to Pending after each successful execution, with the next run time updated accordingly.

Viewing Results

After a task completes, you can view the execution results in two ways:
  1. Notification — Check the notification you received (email, Feishu, or in-app) for a summary
  2. Task Detail — Click on the task in the task list to see the full Agent session, including all messages, tool calls, and outputs
Each execution creates a dedicated session, so you have a complete history of every run.

Managing Tasks

From the Automated Tasks page, you can:
  • Edit — Update the prompt, schedule, or notification settings of an existing task
  • Pause / Resume — Temporarily suspend a task without deleting it
  • Delete — Permanently remove a task and its schedule
  • View History — See past execution results, run count, and last/next run times

Use Cases

ScenarioExample PromptSchedule
Daily Standup Summary”Summarize yesterday’s GitHub commits and open PRs for our team”Daily, 9:00 AM
Weekly Report”Generate a weekly progress report based on completed Jira tickets”Weekly, Monday 10:00 AM
Data Monitoring”Check the error rate dashboard and alert me if it exceeds 5%“Daily, every 6 hours
Content Publishing”Draft and publish today’s social media post based on our content calendar”Daily, 8:00 AM
Code Review Reminder”List all PRs that have been open for more than 48 hours without review”Daily, 2:00 PM

Architecture

The automation system is built on a two-component architecture:
ComponentResponsibility
Web (Next.js)Task CRUD, user authentication, result storage
AI Backend (Python)Schedule management, Agent execution, notification delivery
Reliability: When the AI Backend restarts, it automatically syncs all pending tasks from the Web API, ensuring no scheduled tasks are lost.