Custom Actions is available on Pro Plus.
How It Works
- Define an action — give it a name, description, endpoint URL, and parameters
- The AI learns when to use it — based on your description
- During a conversation, the AI calls your endpoint and uses the response to help the customer
Creating an Action
| Field | Description |
|---|---|
| Display Name | Human-readable name (e.g. “Check Inventory”) |
| Action Name | A short identifier for the action |
| AI Description | Tell the AI when and why to use this action |
| Endpoint URL | Your webhook URL (must be HTTPS) |
| HTTP Method | GET or POST |
| Parameters | Named parameters the AI will extract from the conversation |
Parameters
Each parameter has:- Name — A descriptive name for the value
- Type — String, Number, or Boolean
- Description — Helps the AI understand what to extract from the conversation
- Required — Whether the AI must have this value before calling
Example: Check Inventory
Say you want the AI to check part availability. You’d create an action like:- Name: Check Inventory
- Description: Check if a specific part is in stock. Use when a customer asks about part availability.
- Endpoint:
https://api.yourshop.com/inventory/check - Parameters: Part name and device model
Response Format
Your endpoint should return JSON. The AI will interpret the response and present it conversationally. There’s no required schema — the AI is flexible enough to handle most JSON structures.Security
- Endpoints must use HTTPS
- Requests are time-limited to prevent slow endpoints from blocking conversations
- Call counts are tracked per action for monitoring
- Actions can be toggled active/inactive without deleting
Use Cases
- Inventory checks — Real-time stock availability
- CRM lookups — Check customer history or loyalty status
- Warranty validation — Verify warranty status by serial number
- External pricing — Query a third-party pricing API
- Workflow triggers — Kick off processes in Zapier, Make, or n8n