The Agent Capture Node

The Agent Capture Node is designed to extract variables from a user during a conversation — seamlessly and conversationally. It allows your AI Agent to gather key information like order numbers, contact details, or custom fields before moving forward in the flow.

The true power of the Agent Capture Node lies in its ability to collect variables naturally as part of a guided interaction — not just by asking one question, but by holding a short, goal-oriented conversation.

For example, when a user triggers the “Where’s my order?” intent, the Agent Capture Node can be used to gather a 6-digit order number before moving on to the next step.

Common use cases include:

• Collecting lead information

• Interviewing job candidates and storing responses

• Gathering customer feedback

• Capturing context like email, symptoms, or policy numbers

Once the required variables are captured, the conversation continues to the next node — such as an API Call Node that fetches the order status, followed by a Say node that shares the result.

You can also use the Agent Capture Node to handle short interactions when only one variable is required after an API call.

For instance, if an API returns data such as "delivery_status": "Delivered" and "delivery_time": "2:30 PM", your Agent Capture Node can use this information (via {last_api_response}) to respond:

“Great news! Your order was delivered today at 2:30 PM. Let me know if there’s anything else I can help you with.”

💡 You can also use a Say Node for similar output-only responses.

Note: you can also use the "Say" node for this purpose.

This is where you define how your AI Agent should behave during this variable collection step.

You can specify:

• Goal: What the Agent should achieve in this interaction

• Instructions: Step-by-step guidance (e.g., “Ask for the customer’s 6-digit order number”)

• Role: Define the Agent’s persona (e.g. Customer Support Agent)

• Response Style: Tone and manner (e.g., clear and professional)

• Limitations: What the Agent should do when uncertain

• Character: High-level personality or behavior traits

💡 Example:

• Goal: Help users locate their order

• Instructions: Ask for their 6-digit order number. If they don’t have it, explain how to find it under “profile > my orders.”

• Role: Support representative

• Style: Friendly and direct

• Limitation: Say “I’m not sure” if you can’t proceed, and offer to escalate

• Character: Calm, helpful, professional

Use { to insert variables into your instructions. For example:

“Use the order number from {order_id} to retrieve delivery details.”

To learn more about instructions, click here.

Here, you choose the AI model. Different models have different strengths, so select the one that fits your needs. GPT4o, GPT4o Mini and GPT3.5 are available on all plans. GPT4 and GPT4-turbo are available only on paid plans.

This controls how creative or deterministic the AI's responses should be. A higher temperature means more creative responses, while a lower temperature makes responses more predictable and accurate.

You have full control over when the Agent Capture Node should exit and pass the conversation forward. Choose from:

1. All Variables Captured (Default)

The node will only exit once all required variables are successfully captured. Ideal for forms or onboarding steps.

2. Any Variable Captured

The node exits once at least one of the required variables is captured. Great when partial data is enough to proceed.

3. Continue Until Exit is Triggered

The Agent will stay in the node and continue chatting until another flow is triggered or an exit condition is met. Use this for extended interactions or follow-ups.

This tab lets you define which variables the Agent should extract during the conversation.

🔄 Capture Behavior

You can control how frequently each variable is captured:

• Always Capture: The variable is always collected (even if it exists).

• Capture on Every Flow Execution: The variable resets when the flow is re-entered.

• Capture Once per Conversation: If the variable is already collected earlier in the conversation, it won’t be asked again.

🧠 Pro Tip:

Include variable instructions in your node like:

“Greet the user, ask for their full name, then ask for their order number, and finally ask them to describe their issue.”

In the Conversation Audit, you can review:

• The AI’s instructions

• Which variables were captured

• How the Agent phrased its questions

• When the node exited

• What tool or logic was triggered next

This visibility helps you refine prompts and understand where users might drop off or need clarification.

The Agent Capture Node is ideal for collecting structured data in a conversational way, ensuring your workflows stay clean while your users stay engaged.