Tool★

A tool (also called a function or capability) is an executable function exposed to an LLM by the client harness. It lets the model interact with the external world (e.g. read files, run terminal commands, query databases, or call web APIs).

You define tools whenever you:

• Write an integration to let an agent search Google.
• Connect a model to your custom database API.
• Configure file readers/writers for an IDE assistant.

Tool Definition via JSON Schema

AI models do not execute code natively. To tell the model that a tool exists, you must define its interface in the model request payload using a JSON Schema. This tells the model:

1. The Name of the tool (e.g. calculate_sum).
2. A Description of what it does (e.g. "Calculate the sum of two integers").
3. The Parameters it accepts, their types, and which fields are required.

The model reads these schemas and uses them to determine when and how to call the tool.

Field Applications & Schema Construction

1. Fullstack Developers (Defining Tool Schema in Request)

• Code Example:

  const toolDefinitions = [
    {
      name: "search_web",
      description: "Search the web for current information.",
      input_schema: {
        type: "object",
        properties: {
          query: {
            type: "string",
            description: "The search query (e.g. 'Astro framework version')."
          }
        },
        required: ["query"]
      }
    }
  ];