Guide FlowsAI Mode

AI Mode

AI Mode lets an AI model automatically generate missing parameters for your tools and utilities, making it easier to configure complex API calls without manually specifying every field.

How AI Mode Works

When you enable AI Mode on a tool node, you provide a natural language instruction describing what you want to accomplish. The AI then generates the appropriate parameters for any empty fields.

You Provide

Instruction:

"Create a new lead named John Smith with email john@example.com"

Empty fields left blank

AI Generates

{
  "first_name": "John",
  "last_name": "Smith",
  "email": "john@example.com"
}

Only Empty Fields

AI only generates values for fields you leave empty. Any fields you fill in yourself are kept as-is.

Schema-Aware

AI understands the tool's schema and generates properly structured data matching required types and formats.

Enabling AI Mode

Tool Configuration Panel

AI Mode

AI Model

claude-sonnet-4-5-20250514

Model used to generate missing parameters

Instruction for AI

Tell AI what you want to accomplish...

Describe what you want to accomplish. AI will generate values for empty fields.

Process Response with AI

Convert raw response to human-readable format

AI Mode Settings

AI Model

Select which AI model to use for parameter generation. Different models have different capabilities and costs.

Claude Models:

• Claude Sonnet (Recommended)
• Claude Opus (Most capable)
• Claude Haiku (Fastest)

OpenAI Models:

• GPT-4o
• GPT-4o Mini
• o1 / o3 (Reasoning)

Instruction for AI

Describe in natural language what you want the tool to do. Be specific about the data you want to send or receive.

Good Examples:

• "Create a contact with first name from trigger and email john@company.com"
• "Search for all opportunities with status 'open' in the last 30 days"
• "Send a POST request to create a new user with the data from the previous step"

Use Template Values:

You can use template syntax in your instruction: "Create contact with name {{#1.name}}"

Process Response with AI

When enabled, AI will convert the raw tool response into a human-readable format, summarizing key information and explaining what happened.

Raw Response:

{"id":"cust_123",
"created":1702000000,
"livemode":false,...}

AI Processed:

"Successfully created customer with ID cust_123 in test mode. The customer was created on December 8, 2023."

Visual Indicators

When AI Mode is enabled, visual indicators help you understand what values were generated by AI versus what you provided.

Execution Input Display

{
  "toolParameters": {
    "first_name": "John",
    "last_name": "Smith",
    "email": "john@example.com",
    "phone": "+1-555-0123"
  },
  "aiMode": {
    "enabled": true,
    "model": "claude-sonnet-4-5-20250514",
    "instruction": "Create lead with name John Smith...",
    "processResponse": false,
    "aiGeneratedParameters": {
      "last_name": "Smith",
      "phone": "+1-555-0123"
    },
    "usage": {
      "inputTokens": 245,
      "outputTokens": 42,
      "totalTokens": 287,
      "totalCredit": 0.000287
    }
  },
  "userInput": {
    "first_name": "John",
    "email": "john@example.com"
  }
}

toolParameters

Final merged values sent to the tool

aiGeneratedParameters

Values that AI generated for empty fields

userInput

Values you provided directly

Usage & Credits

AI Mode Uses API Credits

Each time AI Mode generates parameters or processes a response, it makes an API call to the selected AI model. This uses your API credits based on the token count.

Token usage is tracked:

• inputTokens - Tokens in the schema and instruction
• outputTokens - Tokens in the generated parameters
• totalCredit - Total credit cost for this operation

Best Practices

Do This

• Be specific in your instruction about what data to use
• Pre-fill known values - AI only generates for empty fields
• Use template syntax for dynamic values from previous nodes
• Enable Process Response only when you need readable output
• Choose faster models (Haiku/GPT-4o Mini) for simple parameter generation

Avoid This

• Vague instructions like "do something with this data"
• Overusing AI Mode when you know the exact values needed
• Complex nested logic - AI works best for straightforward parameter mapping
• Large models for simple tasks - wastes credits and is slower
• Assuming AI knows context - always provide explicit data references

When to Use AI Mode

💡

Complex API Schemas

When a tool has many fields and you only need to set a few based on natural language description.

🔄

Data Transformation

When you need to map data from one format to another and AI can infer the correct mapping.

📝

Response Summarization

Use "Process Response" to get human-readable summaries of complex API responses.

When NOT to Use AI Mode

• When you know exactly what values to use - just fill them in directly
• For simple field mapping - use template syntax like {{#1.field}}
• In performance-critical flows - AI adds latency
• For highly sensitive data - values go through an AI model

Example: CRM Lead Creation

Scenario: Creating leads from webhook data

Without AI Mode

You need to manually map each field:

first_name: {{#1.first_name}}

last_name: {{#1.last_name}}

email: {{#1.email}}

phone: {{#1.phone}}

company: {{#1.company}}

source: "Website Form"

status: "New"

With AI Mode

Just describe what you want:

"Create a new lead using all the contact information from the webhook trigger. Set the source as 'Website Form' and status as 'New'."

AI will automatically map matching fields and set the specified values.

Advanced Flow Control How Flows Work