Skip to Content
ProductsWorkflow BuilderWorkflowNodesAgent

Agent Node — Settings

The Agent node is responsible for generating responses and performing actions within a workflow.
It defines how the AI behaves, what model it uses, how it remembers context, and whether it can interact with external tools.

agent Node

This document explains each setting clearly and conceptually so anyone can understand how to configure an Agent node.

Agent Configuration Overview

agent Node settings

When you add an Agent node, you configure:

  • Identity and behavior
  • Model selection and response style
  • Output structure (optional)
  • Memory handling
  • Tool integrations
  • Streaming and answering behavior

Name

Purpose
Identifies the agent within the workflow.

Guidelines

  • Use a clear, descriptive name (e.g., Support Assistant, Partner Rewards Agent).
  • The name helps distinguish this agent from others in complex workflows.

System Prompt

Purpose
Defines the role, personality, and behavior of the agent.

What to include

  • The agent’s responsibility
  • Tone or style (formal, friendly, concise, etc.)
  • Any high-level rules or boundaries

Example (conceptual)

“You are a customer support assistant helping users understand partner reward programs.”

This prompt guides how the agent thinks and responds, not what nodes do next.

Provider

Purpose
Selects the AI provider used by the agent.

Behavior

  • Determines which platform supplies the language model.
  • Available providers depend on your workspace configuration.

Model

Purpose
Chooses the AI model used to generate responses.

Impact

  • Different models may vary in speed, accuracy, and cost.
  • Model choice affects how detailed or concise responses are.

Temperature (0.0 – 2.0)

Purpose
Controls how creative or deterministic the agent’s responses are.

How it works

  • Lower values (e.g., 0.2–0.5) → More consistent and predictable responses
  • Higher values (e.g., 1.0–2.0) → More varied and creative responses

Typical use

  • Support or compliance flows → Lower temperature
  • Brainstorming or ideation → Higher temperature

Max Tokens

Purpose
Limits the maximum length of the agent’s response.

Behavior

  • Prevents overly long outputs.
  • Helps control response size and cost.

Output Schema (Optional)

agent Node Output Settings

Purpose
Defines a structured format for the agent’s response.

When to use

  • When downstream nodes expect structured data.
  • When consistency in output format is required.

Output Properties

Each property includes:

  • Name – Identifier of the field
  • Type – Data type of the value
  • Description – What the value represents

Supported Types

  • STR – String
  • NUM – Number
  • BOOL – Boolean
  • ENUM – Predefined set of values
  • OBJ – Object
  • ARR – Array

If no output schema is defined, the agent responds in free text.

Enable Memory

Purpose
Allows the agent to remember information from earlier messages.

Behavior

  • When enabled, the agent can maintain conversational context.
  • When disabled, each interaction is handled independently.

Use Case

  • Enable for multi-step conversations.
  • Disable for one-time or stateless interactions.

Tools Configuration

agent Node tools Settings

Purpose
Extends the agent’s capabilities by allowing it to interact with external systems or knowledge sources.

Adding a Tool

Click Add Tool to connect external services.

Tool Type: API Request

agent Node API tools Settings

Allows the agent to call an external API during workflow execution.

API Request Settings

Endpoint URL
  • The external API endpoint the agent can call.
Tool Name
  • A short identifier used internally by the agent.
HTTP Method
  • Determines how the request is sent (e.g., GET, POST).
Retry (Optional)
  • Number of retry attempts if the request fails.
Description
  • Explains what the API does and when the agent should use it.
Send Query Parameters

agent Node Send Query Settings

Purpose

  • Adds query parameters to the API request.

Fields

  • Parameter Name
  • Parameter Value
Send Headers

agent Node Send Headers Settings

Purpose

  • Adds custom headers to the API request.

Fields

  • Header Name
  • Header Value
Send Body

agent Node Send Body Settings

Purpose

  • Sends structured data in the request body.

Body Parameter Types

  • String
  • Number
  • Boolean
  • Object
  • Array

Value Source

  • Static
  • Dynamic
Enable Authorization

agent Node Authorization Settings

Purpose

  • Secures API requests using authentication.
Supported Authorization Types
Bearer Token
  • Uses a token for authorization.
Basic Auth
  • Uses username and password.
API Key
  • Sends an API key via a header.
Custom Header
  • Uses a custom header name and value

Tool Type: Knowledge Base Retrieval

agent Node Knowledge tools Settings

Allows the agent to answer questions using an internal knowledge base.

Knowledge Base Settings
Knowledge Base
  • Selects which knowledge base the agent can access.
Tool Name
  • Identifier used by the agent.
Description
  • Explains how the agent should use this knowledge base.
Similarity Threshold (0.0 – 1.0)
  • Controls how strict matching is when retrieving content.
    • Lower values → More flexible matches
    • Higher values → More precise matches

Enable Streaming

agent Node Enable Streaming Settings

Purpose
Controls whether the agent sends responses incrementally.

Behavior

  • When enabled, users see responses as they are generated.
  • When disabled, the response appears all at once.

Enable Answering

agent Node Enable Answering Settings

Purpose
Determines whether the agent is allowed to respond to the user.

Behavior

  • When disabled, the agent may perform actions without producing a visible response.

Summary

The Agent Node allows you to:

  • Define AI behavior and tone
  • Choose models and response style
  • Structure outputs for downstream logic
  • Maintain conversational memory
  • Integrate APIs and knowledge bases
  • Control response streaming and visibility

Together, these settings make the Agent node the core intelligence of your workflow.

StartAuto Orchestrator