Creating an Agent

This guide walks you through creating a new AI agent, configuring its skills and permissions, and assigning it to event streams.

Note

You need the Manage Organization permission to create and manage agents.

1. Navigate to the Agents page

   Open the SuiteOp dashboard and select Agents from the sidebar navigation.

2. Click Create Agent

   Click the Create Agent button to open the agent configuration form.

3. Set the basic details

   Fill in the following fields:

   • Name — a short, descriptive name for the agent (e.g., “Temperature Monitor”, “Guest Concierge”). Maximum 100 characters.
   • Description — an optional summary of what the agent does. Maximum 1,000 characters. This is visible in the agents list.
   • System Prompt — custom instructions that define the agent’s personality, tone, and behavior guidelines. Maximum 10,000 characters. This is the primary way you control how the agent thinks and responds.

4. Configure skills

   Skills are ordered blocks of instructions that compose the agent’s capabilities. Each skill has:

   • Name — a label for the skill (e.g., “Temperature Management”)
   • Description — a short summary of what the skill does (up to 500 characters)
   • Instructions — detailed instructions for the agent when this skill applies (up to 5,000 characters)
   • Enabled — toggle the skill on or off without deleting it
   • Property scope (optional) — restrict the skill to specific properties or property groups. If no scope is set, the skill applies to all properties.

   You can add up to 20 skills per agent. Skills are evaluated in order, so place the most important ones first.

   Note

   Skills with property scoping are only activated when the agent processes events from matching properties. This lets you give different instructions for different property types — for example, stricter temperature rules for luxury properties.

5. Set tool permissions

   Tool permissions control what the agent is allowed to do. Each tool can be set to one of three modes:

   Mode	Behavior
   Auto	The agent can use this tool without asking for approval. Actions are logged but executed immediately.
   Ask	The agent proposes the action but waits for a human to approve or reject it before execution.
   Disabled	The agent cannot use this tool at all.

   See Agent Tools & Actions for the full list of available tools and recommendations for each permission mode.

6. Set budget limits

   Configure the token limits to control costs:

   • Daily token limit — defaults to 5,000,000 tokens. The agent pauses if it exceeds this in a calendar day.
   • Monthly token limit — defaults to 50,000,000 tokens. The agent pauses if it exceeds this in a calendar month.

7. Set the batch window

   The batch window controls how long the system waits to collect events before processing them as a group. The default is 5 seconds, with a range of 1 to 60 seconds.

   A shorter window means faster response times but potentially more sessions (and token usage). A longer window batches more events together, which can be more efficient.

8. Save the agent

   Click Save to create the agent. It will be created in the Active status by default.

9. Assign event streams

   After creating the agent, assign it to event streams so it starts receiving events:

   1. Go to the agent’s detail page
   2. In the Stream Assignments section, click Assign Stream
   3. Select the event type you want this agent to handle
   4. Click Assign

   Repeat for each event type you want the agent to monitor.

Caution

Before assigning an agent to live event streams, use the sandbox to test its behavior with simulated events. This lets you verify the agent responds correctly without affecting real data.

Editing an Agent

You can update any agent configuration at any time from the agent’s detail page. Changes to tool permissions automatically update the agent’s organization membership permissions to stay in sync.

Pausing and Deleting Agents

• To temporarily stop an agent, change its status to Paused. Paused agents do not process live events but can still be used in the sandbox.
• Deleting an agent is a soft delete — it sets the agent’s status to Paused rather than removing the record. This means a deleted agent can be restored by changing its status back to Active. All session history and action logs are preserved for audit purposes regardless of whether the agent is paused or deleted.