Give the agent actions it can run: search data, capture leads, book meetings, take payments, and move visitors to the right page.
The AI chatbot tools and actions reference explains every action you can attach to an agent. Tools extend what the chatbot can do beyond answering from trained content: search a database, look up a booking, capture a lead, send a transcript, open a payment flow, or navigate a visitor to another page. Each guide covers configuration, required credentials, and how to test the action before it goes live.
Set up tools from the Tools and Actions overview. An action is a call the agent can make when a conversation needs live data or a side effect. You define when it can run, which fields it collects, and what it returns to the conversation. Start with one action, test it with a real question, then add the next.
Use these actions when the answer lives outside the trained content.
Let the agent check availability and book a meeting without leaving the conversation.
Open the action, confirm the credentials or connection, then ask a question that should trigger it. A passing test returns the expected value in the conversation and leaves a clear trace you can inspect. If the action does not fire, check the description, the required inputs, and whether the model can see the fields it needs. Disable an action before changing its schema, and re-test after every edit.
