Skip to main content

​
Overview

The Agent Node is a powerful integration node that connects your workflows to AI language models for text generation, data processing, and intelligent automation.

​
Required Configuration

​
Basic Settings

  • Label - Display name for the node
  • Agent Name - Unique identifier for the agent
  • Instructions - Behavior definition and system prompt
  • Agent Input Configuration - Configure how the agent receives input data using ‘Fixed’ for static text, ‘From args’ for field references, or ‘Expression’ for dynamic values

​
Model & Operation Settings

  • Operation – Select one: generateText, generateObject
    • generateText: Designed to produce natural, conversational language, such as a chat response or a summary for a human reader. Use generateText when you want the AI to talk to your users.
    • generateObject: Built to output structured, machine-readable JSON that your code can reliably parse and use for logic or database storage. Use generateObject when you need the AI to behave like an API that returns specific, typed data fields.
  • Base URL (Optional) – For OpenAI-compatible APIs (defaults to https://api.openai.com/v1). Change this to connect to other providers such as Anthropic, Google, Groq, etc.
  • API Key – Your API key from your chosen provider. For OpenAI, get it from the link shown in the UI. For other providers, get it from their respective dashboards.
  • Chat Model – Dynamically fetched from your API provider, or you can enter a custom model name.

​
Output Schema

  • Output Schema (JSON) - Define the JSON structure your agent will return (required for generateObject operation)

​
Optional Configuration

  • Requires Auth - Whether this node needs user authentication (default: false)
  • Enable Streaming - Real-time response streaming (default: true)

​
How It Works

​
Step 1: Basic Setup

Configure your agent’s identity, behavior, and input source. The input configuration supports:
  • Fixed values - Static text
  • Field references - Dynamic values from upstream nodes
  • Expressions - Complex value transformations

​
Step 2: Model Configuration

  1. Enter your API key (automatically validated)
  2. Fetch available models from your provider or type a custom model name
  3. Select your operation type (generateText or generateObject)
  4. Enable/disable streaming responses

​
Step 3: Output Schema

For generateObject operations, define a JSON schema to structure your agent’s responses consistently.

​
Model Support

The Agent Node supports ANY OpenAI-compatible API provider. Most modern AI providers offer OpenAI-compatible endpoints, making it easy to connect to virtually any LLM service by simply changing the Base URL and providing the appropriate API key.

​
Supported Providers

Connect to any of these providers (and more!) by configuring the Base URL:
  • OpenAI - Native support for GPT models
  • Google AI - Gemini models via OpenAI-compatible endpoint
  • Groq - Ultra-fast inference
  • Together AI - Open-source models
  • Fireworks - Fast inference platform
  • DeepSeek - Advanced reasoning models
  • Mistral - Mistral and Mixtral models
  • OpenRouter - Access to 100+ models
  • Hugging Face - Thousands of open-source models
  • Novita AI - Multi-model platform
  • xAI - Grok models
  • Azure OpenAI - Enterprise-grade OpenAI hosting
  • Vercel AI Gateway - AI proxy and caching
  • vLLM - Self-hosted model serving
  • And many more…

​
How to Connect

  1. Enter the Base URL for your provider (e.g., https://api.groq.com/v1)
  2. Add your API Key from the provider
  3. Select the model from the list of models fetched dynamically

​
Example Configurations

OpenAI (Default) Groq
  • Base URL: https://api.groq.com/v1
  • API Key: Get from your Groq dashboard
  • Models: llama-3.1-70b, mixtral-8x7b, etc.
OpenRouter
  • Base URL: https://openrouter.ai/api/v1
  • API Key: Get from openrouter.ai
  • Models: Access to 100+ models from various providers

​
Connections

Agent nodes can connect to:
  • Query/Mutation nodes for further processing
  • Frontend Element nodes to output results
  • Other Agent nodes (chained processing)
  • Other transformation nodes

Query Node

Fetch data to process with AI

Mutation Node

Store AI-generated results