What This Feature Does
Your Rep AI assistant can call external APIs during conversations with your customers. This means your AI can pull live data from any service you use — exchange rates, product recommendations from your own systems, inventory checks, appointment availability, or any custom tool your business relies on — and use that information to help shoppers in real time.
Why This Matters for Your Business
Make your AI uniquely yours. If you've built custom tools, recommendation engines, or internal services, your Rep AI assistant can now tap into them directly. Instead of shoppers leaving your chat to check another tool or contact support, your AI handles it in the conversation.
Real-world examples of what you can do:
A beauty brand with a proprietary hair color analyzer has the AI call their matching algorithm and suggest products based on the result
A store selling internationally has the AI check live exchange rates before quoting prices
A merchant with a custom sizing tool has the AI call their API and recommend the right fit
A store with a booking system has the AI check appointment availability and suggest times
How It Works
When a customer asks a question or takes an action that requires external data, your AI assistant automatically calls the API you've configured, retrieves the response, and uses it to answer the customer — all within the same conversation. The customer never leaves the chat.
Your AI supports three types of API calls:
GET — Retrieve information (e.g., check an exchange rate, look up availability)
POST — Send data and get a result back (e.g., submit a quiz answer, run a recommendation)
PUT — Update information in an external system (e.g., save a preference, update a record)
Important: You're Responsible for Your API Integrations
Please read this carefully before setting up an external API connection.
When you connect an external API, your AI assistant will execute real calls to that service during live customer conversations. This is a powerful capability, but it comes with responsibility.
You are fully responsible for the APIs you connect and the actions they perform. This includes the accuracy of the data returned, any changes made to external systems through POST or PUT calls, and any costs or rate limits associated with your API usage.
External API calls may interact with other Rep AI features in unexpected ways. For example, if your API returns product recommendations, they may overlap or conflict with Rep AI's built-in product suggestions, sales skills, or catalog-based responses. If your API modifies cart contents or applies discounts, it could interfere with Rep AI's cart management or discount generation features. Test thoroughly to make sure your API integration works well alongside your existing Rep AI setup.
A few things to keep in mind:
Rep AI does not validate, monitor, or guarantee the behavior of third-party APIs you connect
If your API goes down or returns errors, the customer experience may be affected — always include fallback instructions for your AI
Sensitive data sent to or received from your API is your responsibility to manage in accordance with your own privacy policies
API response times directly affect how quickly your AI replies — slow APIs mean slower conversations
When in doubt, start with read-only GET requests before enabling POST or PUT calls that modify data
We recommend testing any new API integration extensively in Test & Train before going live.
Setting It Up
To connect an external API, you'll provide your AI with instructions that include the API details. Here's how:
Step 1: Prepare your API information
Before configuring, make sure you have:
The API endpoint URL (e.g.,
https://api.yourservice.com/v1/recommendations)The HTTP method (GET, POST, or PUT)
Any required parameters (URL parameters or request body fields)
Authentication credentials if needed (API key/token, or username and password)
Step 2: Add instructions in your AI settings
Go to AI Training > AI Personality in your Rep AI Console
In the General AI Instructions field, describe when and how the AI should call your API
Include the endpoint URL, method, parameters, and authentication details
Explain what the AI should do with the response — how to interpret it and what to tell the customer
Click Save
Step 3: Test the integration
Go to AI Training > Test & Train
Simulate a conversation where the API call would be triggered
Verify the AI calls your API correctly and presents the results naturally
Adjust your instructions if the response needs refinement
Writing Effective API Instructions
The quality of your instructions directly affects how well your AI uses the API. Here are some tips:
Be specific about when to call the API. Tell the AI exactly which customer requests or situations should trigger the call.
Example:
"When a customer asks about currency conversion or mentions paying in a different currency, call the exchange rate API at https://api.exchangerate-api.com/v4/latest/USD using a GET request. Use the returned rates to convert the product price and show both the original USD price and the converted amount."
Describe the expected response. Help your AI understand what the API returns and how to use it.
Example:
"The API returns a JSON object with a 'rates' field containing currency codes and their exchange rates relative to USD. Use the customer's local currency based on their location to calculate the converted price."
Include error handling. Tell the AI what to do if the API call fails.
Example:
"If the exchange rate API is unavailable, let the customer know that prices are listed in USD and suggest they check a currency converter for an estimate."
Provide authentication details clearly.
Example:
"Use the API key Bearer YOUR_API_KEY_HERE in the Authorization header when calling the endpoint."
Best Practices
Start simple. Begin with one API integration and expand from there once you've verified it works well
Test thoroughly. Use Test & Train to simulate different customer scenarios before going live
Keep your API reliable. Your AI depends on the external service being available — choose stable, well-maintained APIs
Protect sensitive data. Never ask the AI to share raw API keys or internal system details with customers
Monitor conversations. Check your Conversations page regularly to see how the AI handles API responses in real interactions
Update instructions when APIs change. If your API endpoints, authentication, or response format change, update your AI instructions immediately
Troubleshooting
The AI isn't calling my API
Check that your instructions clearly describe when the API should be called
Verify the endpoint URL is correct and accessible from the internet
Make sure the API method (GET/POST/PUT) matches what your service expects
Test the API independently to confirm it's working
The AI calls the API but gives wrong information
Review how you've described the API response in your instructions
Add more detail about which fields to use and how to present them
Check if the API response format has changed
The API call is slow and the customer waits too long
Ensure your API responds quickly (under a few seconds is ideal)
Consider whether the data could be provided a simpler way if speed is an issue
Authentication errors
Double-check your API key, token, or credentials in the instructions
Verify the authentication method matches what your API expects (Bearer token, Basic auth, etc.)
Check if your API key has expired or been rotated