Custom mid-call tools are one of the most powerful features in Live Bots 365. They allow your AI assistant to reach out to any external system during a live phone call — looking up data, triggering actions, or posting information — all without pausing the conversation in a way that feels unnatural to the caller.
A mid-call tool works by making an HTTP request to a URL you specify. That URL can be your own server, a third-party API, a webhook endpoint, or a no-code automation platform like Zapier or Make. The assistant sends the request, receives a response, and can use the response data to continue the conversation intelligently.
Creating a Custom Tool
Navigate to Mid-Call Tools in your dashboard and click Create Tool. You will be presented with a configuration form with the following fields:
A short, descriptive name for the tool. This is used internally and also helps the AI understand when to invoke the tool. Use clear names like 'send_sms_link' or 'check_order_status' rather than generic names like 'tool1.'
A plain-language explanation of what the tool does and when the assistant should use it. This is the most important field — the AI reads this description to decide when to call the tool. Be specific: 'Use this tool when the caller asks to receive a link via text message' is far more effective than 'sends a text.'
The full URL the tool will call when invoked. This must be a publicly accessible HTTPS endpoint. If you are using a webhook service like Zapier, paste the webhook URL here.
The HTTP method to use for the request — typically POST for actions that send data (like booking or sending SMS) or GET for lookups (like checking availability).
Any HTTP headers required by the API, such as Authorization tokens or Content-Type. Add each header as a key-value pair. For bearer token authentication, add a header with key 'Authorization' and value 'Bearer YOUR_TOKEN'.
The data fields the assistant will send with the request. For each parameter, you define a name, a type (string, number, boolean), and a description. The AI uses the description to understand what value to extract from the conversation and pass to the tool.
Understanding Parameters
Parameters are the pieces of information the assistant extracts from the conversation and sends to your API. For example, if you are building a "Send SMS" tool, you might define a parameter called to with the description "The caller's phone number in international format." The AI will automatically extract the caller's phone number from the conversation context and pass it as the value for this parameter.
You can also define parameters with fixed values — for example, a from parameter that always uses your business phone number ID. In this case, describe the parameter as "Always use the value 7786" and the AI will consistently pass that value without trying to extract it from the conversation.
Example: Send SMS with Website Link
Here is a real-world example of a custom mid-call tool that sends an SMS message with a website link to the caller during the call:
to (string) — The caller's phone number in international format (e.g., +15551234567)from (number) — Always use the value 7786 (your SMS-capable phone number ID)body (string) — Always use: "Here is the link to Live Bots 365 that you requested. livebots365.com — we are excited to activate your account."Assigning Tools to an Assistant
After creating a tool, you need to assign it to the assistants that should have access to it. Open the assistant's configuration page, scroll to the Tools section, and toggle on the tools you want that assistant to be able to use. An assistant can only invoke tools that have been explicitly assigned to it.
You can assign the same tool to multiple assistants, which is useful when you have a common action — like sending a follow-up SMS — that should be available across several different agents.
Security Note
Never expose sensitive credentials directly in tool parameter values. Use your API's authentication headers (such as Bearer tokens) to secure the connection. If your API requires a secret key, add it as an Authorization header rather than passing it as a parameter that the AI might inadvertently expose in a conversation.