AI Agent Tool

Node catalog Artificial intelligence

On this page

Nested agent as a tool

Usage and specifics

Select the published workflow and its accessible agent, then build its input. Target agent permissions and limits still apply. Cycles are rejected; tool approval cannot pause a nested agent.

Technical type : ai_tool_ai_agent. Runtime : nodulz.

Credentials and prerequisites

This node defines no dedicated external credential. Remote sources or execution targets may still need their own access credentials.

See the credentials guide if a remote source or execution target requires access.

Parameters

Names below are technical keys. Defaults are not a ready-to-run configuration: complete fields required by your operation. Paths containing [] describe list items; some fields depend on the selected mode.

Parameter	Type	Default	Values / limits
`workflowId`	`string`	`""`	-
`targetScope`	`enum`	-	`local`, `workflow`
`name`	`string`	`"nested_agent"`	-
`description`	`string`	`"Nested AI agent"`	-
`agentNodeId`	`string`	`""`	-
`maxDepth`	`number`	`1`	-; ≥ 1, ≤ 3

Configuration help

workflowId Agent workflow: Only published workflows you can read and execute are listed. Permissions are checked on every call.
agentNodeId Agent to call: Choose an enabled AI agent in the workflow. Its answer is available in text, its steps in steps, and its turn count in iterations.
name Tool name: Unique name the orchestrating agent uses to call this tool. Prefer a short name without spaces.
description When to delegate to this agent: Describe its expertise and the tasks to delegate. The orchestrating agent uses this description to choose the tool.
maxDepth Delegation levels: Limits nested agent calls independently of conversation turns. Recursive calls are always rejected.

Initial configuration

Initial catalog configuration example, to adapt before running. Keep secrets in credentials.

json

{
  "workflowId": "",
  "name": "nested_agent",
  "description": "Nested AI agent",
  "agentNodeId": "",
  "maxDepth": 1
}

Verification

Test with representative data. Check node output in the inspector and logs, then use editor mapping for downstream fields. A trigger test with a simulated payload does not validate external service delivery.