How to Create an AI Agent

TL;DR

Space admins can create AI agents under Space Settings → AI Agents by choosing a prebuilt template or describing a custom agent’s purpose in chat. The chat builder sets up its scope, triggers, and actions, which you can refine in chat or the synced manual form. Test the agent in the Playground without using your AI usage allowance, then click Create. For an agent to work, its scope must be defined - either in the agent builder or by @mentioning it in a task, project, or folder.

• Overview
• Build an Agent by Describing It
• How to Create an AI Agent
• What’s Next?

Overview

You can choose from prebuilt agents or create custom ones to suit your needs. Prebuilt agents (Risk, Triaging, and Intake) come with predefined logic so you can deploy them quickly. To create or manage agents, you need Space admin permissions.

Build an Agent by Describing It

You can create a custom agent just by describing what you want it to do in a chat conversation. Tell the builder the situation to watch for and the action to take, and it sets up a working agent - trigger, scope, and actions - from your description. This is the default way to create a custom agent.

You can keep refining the agent in the same conversation until it behaves the way you want, and you can ask the builder to add a folder, project, or task to its scope so you don't have to @mention it manually afterwards.

Example: You can add this prompt to the chat: "When a new task is added to this folder and it's missing a due date, post a comment asking the requester to add one," and the builder creates an agent configured to watch that folder and post the comment.

How to Create an AI Agent 

1. Navigate to the space where you want to create an agent, then click the Settings 1 icon next to the space overview in the sidebar or below the space title in the overview.

   

2. Select the AI agents 2 tab in the space settings overview and click Get started 3.

   

3. You can either select a ready-made template (Revision Tracker Agent, Project Coordinator Agent, Missed Deadline Agent, Triaging, Intake, or Risk Agent) 4, or use the Chat feature 5  to build a Custom AI agent.

   The chat complements the form without duplication, syncing changes live so you can tune and test without leaving the screen.

4. You can configure the agent's settings 6 and specify the following details:
   1. Name your agent - this serves as its @handle when adding locations to its scope.
   2. Agent scope - Select the tasks, projects, or folders that define this agent's scope. The agent will monitor triggers and take action only within these items and their subitems. 
   3. General instructions - Define the agent’s core purpose and behavior. This guides how it executes the specific actions below.
   4. Set the trigger - Choose the event that activates the agent and where to monitor for it. See Agent Triggers Reference.
   5. Define actions - Set what the agent should do when triggered. Drag to reorder. The agent can use outputs from one action to perform the next. See Agent Actions Reference.
5. Use Test 7 feature to test your setup in the Testing Playground. 
   • Playground runs do not count toward your AI usage limits.

6. Click Create 8  to activate your agent.

   

To activate your agent, add a task, project, or folder to its scope. Agents need a designated scope to function. You can set this up during agent configuration or later under Space Settings > AI Agents. Agents without scope will be saved, but they remain inactive until work items are added to its scope.

What’s Next?

• Appoint an Agent to a Location