Skip to main content

Run an agent tool

Configure one AI Agent to call another and check which inputs and results pass between them.

Use Run an agent to delegate a task to another AI Agent. For example, a purchasing assistant can pass a request ID to a status agent and use the returned status to answer the user.


Accessing the Tools Page

Open the AI Agent that will call the other agent. Select Tools in the top navigation. From here you can add, edit, enable, disable, or delete tools.


Adding the “Run an agent” Tool

  1. Select Add a tool → Run an agent.

  2. In Choose application, select the application that contains the target agent. In Select an AI agent, choose the agent to invoke.

  3. Give the tool a descriptive Name. Choose Add to create it, or Add and configure to set up inputs and outputs.


General Settings

Under Details, confirm the tool name, description, type, application, and selected AI Agent. Set Credentials to use to End user credentials for the caller’s access or Specific user for the selected user’s access. AI Agent settings


Inputs – What is passed to the target AI Agent

The input rows depend on whether the target agent defines input variables. Open Inputs after selecting the target agent.

Case 1: The target AI agent has no input variables

The tool has a single fixed User message input. This is the natural-language instruction passed to the target agent. In this mode the input list contains only this row.

Case 2: The target AI agent has defined input variables

If the target agent has defined one or more input variables:

  • All defined Agent Input Variables are automatically listed as inputs

  • These inputs represent structured parameters the called agent expects

  • You may add additional inputs by clicking + Add input

Each input row includes:

Column

Description

Input name

The variable name expected by the called agent

Input type

The variable’s data type (Text, Number, Boolean, etc.)

Fill using

AI-generated value or Custom value

Value

Required only when using a Custom value

Customize input values

For Dynamically fill with AI, select Customize to set the name and description the AI uses. For example, describe a text variable called Request ID as “The exact purchase request ID supplied by the user.” Compare the input options

Input configuration example. Pass the Text variable Request ID to the target agent. Use Customize to describe the ID when filling with AI.


Completion

Under After running, select Waiting for response if the calling agent needs the result. Choose the option to continue without waiting when the caller can proceed independently.


Outputs available to the agent and other tools

Select + Add output, then set the output name and type. Use Customize to map its value to a variable, message, or result from the target agent. Choose a type that matches the returned value.

Outputs are optional

The tool can run the selected agent with the output list empty. Add output mappings when the caller or a later tool needs to use returned data. You can remove and re-add mappings later.


Summary

  1. In a test application, use a target agent with a Request ID input and a known status for a test request. Submit that ID and check that the target receives the correct value.

  2. Map the returned status to a Text output, choose Waiting for response, and verify that the caller uses the expected status. Submit a request with the ID missing and check how the calling agent handles the missing value.

  3. If the target accepts only User message, run a separate test and check that the instruction reaching the target contains the context it needs.

Did this answer your question?