Skip to main content

​
AI Agent

The AI Agent node calls a large language model to generate text or images. It supports any LLM provider, configurable system instructions, tool connections, and multi-modal operations. AI Agent Node

​
Use Cases

  • Generate summaries, classifications, or structured data from upstream inputs
  • Answer questions based on dynamic context from other nodes
  • Generate images from text prompts
  • Use reasoning models for complex decision-making within an orchestration
  • Build autonomous agents that use other nodes as tools

​
Node Reference

​
Parameters

ParameterTypeRequiredDescription
ProviderSelectYesThe LLM provider to use: OpenAI, Anthropic, Google, or Custom
ModelStringYesThe model ID (e.g. gpt-4o, claude-3-5-sonnet-20241022, gemini-1.5-pro)
OperationSelectYesGenerate a text response or Generate an image
CredentialCredentialYesAPI key credential for the selected provider
InstructionsStringNoSystem instructions that guide the AI behavior
PromptStringYesThe user message or input to the model
Reasoning EffortSelectNoLow, Medium, or High — applicable to GPT-5, o-series, and other reasoning models
Service TierSelectNoThe API service tier to use for the request
Max TokensNumberNoMaximum tokens in the response
TemperatureNumberNoSampling temperature (0–2). Higher = more creative.
TimeoutNumberNoMaximum time to wait for a response (in seconds)

​
Providers

ProviderCredential TypeNotes
OpenAIOpenAI API KeySupports GPT-4, GPT-4o, o1, DALL-E image generation
AnthropicAnthropic API KeySupports Claude 3, Claude 4 models
GoogleGoogle AI API KeySupports Gemini 1.5, Gemini 2 models
CustomBearer Token / API KeyUse any OpenAI-compatible API endpoint

​
Operations

​
Generate a text response

Calls the model with system instructions and a user prompt. Returns the generated text response.
FieldDescription
InstructionsSystem instructions that set the model’s persona or task context
PromptThe user message. Supports template expressions (e.g. {{ webhook.body.message }}).
Reasoning EffortControls the depth of reasoning for supported models (GPT-5, o-series).

​
Generate Image

Generates an image from a text description (requires an image-capable model, e.g. DALL-E 3).
FieldDescription
PromptText description of the image to generate
Image SizeOutput image dimensions (e.g. 1024x1024, 1792x1024)
QualityStandard or HD
StyleVivid or Natural

​
Tool Connections (AI Tool Edges)

Connect other nodes to the AI Agent via an AI Tool edge to give the agent the ability to call those nodes during reasoning. The agent autonomously decides when and how to use each tool. AI Agent with Tools Example: Connect an Airtable node and an HTTP Request node as tools. The agent will query Airtable or call the API as needed to answer the user’s request.

​
Output Data

For Generate Text: 
{
  "text": "The primary benefits of AI automation include...",
  "model": "gpt-4o",
  "usage": {
    "promptTokens": 245,
    "completionTokens": 87,
    "totalTokens": 332
  },
  "finishReason": "stop"
}
Reference in downstream nodes: 
{{ aiAgent.text }}
{{ aiAgent.usage.totalTokens }}
For Generate Image: 
{
  "imageUrl": "https://oaidalleapiprodscus.blob.core.windows.net/...",
  "revisedPrompt": "..."
}
  • SubVerse Agents — Run a full multi-turn SubVerse AI agent with channel support
  • Conditions — Branch based on AI output
  • Slack — Send the generated text to a Slack channel