Tool Calling
AI can talk, but how do I get it to actually check the database or make a request?
Can you check whether I’m free tomorrow afternoon?
Request execution based on tool definitions: Tools are defined with a name, description, and typed parameters. The model fills parameters to request a tool call, but this is only an invocation request—the model does not run the tool directly.
The application validates permissions and parameters: It checks whether the tool is permitted, whether parameters make sense, and whether the user has access.
Real execution results return to the model: Only the executed result from the external tool comes back to the model to formulate the final answer.
You can say this to an AI Agent
Validate meeting creation with a mock or dedicated test calendar, not a real schedule. Show the proposed time, attendees, and title before the call; simulate one success and one permission denial, reporting only from the actual tool result. Delete any event created in the test calendar afterward.
Select page elements for precise AI edits
Click a button, text, or card, note the change, then copy its exact location to AI.
Open Selector Pro website↗