Agent Block YAML Schema
YAML configuration reference for Agent blocks
Schema Definition
type: object
required:
- type
- name
properties:
type:
type: string
enum: [agent]
description: Block type identifier
name:
type: string
description: Display name for this agent block
inputs:
type: object
properties:
systemPrompt:
type: string
description: Instructions that define the agent's role and behavior
userPrompt:
type: string
description: Input content to process (can reference other blocks)
model:
type: string
description: AI model identifier (e.g., gpt-4o, gemini-2.5-pro, deepseek-chat)
temperature:
type: number
minimum: 0
maximum: 2
description: Response creativity level (varies by model)
apiKey:
type: string
description: API key for the model provider (use {{ENV_VAR}} format)
azureEndpoint:
type: string
description: Azure OpenAI endpoint URL (required for Azure models)
azureApiVersion:
type: string
description: Azure API version (required for Azure models)
memories:
type: string
description: Memory context from memory blocks
tools:
type: array
description: List of external tools the agent can use
items:
type: object
required: [type, title, toolId, operation, usageControl]
properties:
type:
type: string
description: Tool type identifier
title:
type: string
description: Human-readable display name
toolId:
type: string
description: Internal tool identifier
operation:
type: string
description: Tool operation/method name
usageControl:
type: string
enum: [auto, required, none]
description: When AI can use the tool
params:
type: object
description: Tool-specific configuration parameters
isExpanded:
type: boolean
description: UI state
default: false
responseFormat:
type: object
description: JSON Schema to enforce structured output
required:
- model
- apiKey
connections:
type: object
properties:
success:
type: string
description: Target block ID for successful execution
error:
type: string
description: Target block ID for error handling
Tool Configuration
Tools are defined as an array where each tool has this structure:
tools:
- type: <string> # Tool type identifier (exa, gmail, slack, etc.)
title: <string> # Human-readable display name
toolId: <string> # Internal tool identifier
operation: <string> # Tool operation/method name
usageControl: <string> # When AI can use it (auto | required | none)
params: <object> # Tool-specific configuration parameters
isExpanded: <boolean> # UI state (optional, default: false)
Connection Configuration
Connections define where the workflow goes based on execution results:
connections:
success: <string> # Target block ID for successful execution
error: <string> # Target block ID for error handling (optional)
Examples
Basic Agent
content-agent:
type: agent
name: "Content Analyzer 1"
inputs:
systemPrompt: "You are a helpful content analyzer. Be concise and clear."
userPrompt: <start.input>
model: gpt-4o
temperature: 0.3
apiKey: '{{OPENAI_API_KEY}}'
connections:
success: summary-block
summary-block:
type: agent
name: "Summary Generator"
inputs:
systemPrompt: "Create a brief summary of the analysis."
userPrompt: "Analyze this: <contentanalyzer1.content>"
model: gpt-4o
apiKey: '{{OPENAI_API_KEY}}'
connections:
success: final-step
Agent with Tools
research-agent:
type: agent
name: "Research Assistant"
inputs:
systemPrompt: "Research the topic and provide detailed information."
userPrompt: <start.input>
model: gpt-4o
apiKey: '{{OPENAI_API_KEY}}'
tools:
- type: exa
title: "Web Search"
toolId: exa_search
operation: exa_search
usageControl: auto
params:
apiKey: '{{EXA_API_KEY}}'
connections:
success: summary-block
Structured Output
data-extractor:
type: agent
name: "Extract Contact Info"
inputs:
systemPrompt: "Extract contact information from the text."
userPrompt: <start.input>
model: gpt-4o
apiKey: '{{OPENAI_API_KEY}}'
responseFormat: |
{
"name": "contact_extraction",
"schema": {
"type": "object",
"properties": {
"name": {"type": "string"},
"email": {"type": "string"},
"phone": {"type": "string"}
},
"required": ["name"]
},
"strict": true
}
connections:
success: save-contact
Azure OpenAI
azure-agent:
type: agent
name: "Azure AI Assistant"
inputs:
systemPrompt: "You are a helpful assistant."
userPrompt: <start.input>
model: gpt-4o
apiKey: '{{AZURE_OPENAI_API_KEY}}'
azureEndpoint: '{{AZURE_OPENAI_ENDPOINT}}'
azureApiVersion: "2024-07-01-preview"
connections:
success: response-block