AI Function Calling & Tool Definition Builder

Build type-safe tool schemas and Zod models for OpenAI, Claude and the Vercel AI SDK.

AI Function & Tool Definition Builder

Generate type-safe tool schemas for OpenAI, Anthropic Claude and the Vercel AI SDK.

Preset:

Function Name (snake_case)

Function Description (LLM Instruction)

Parameters (2)

Required

import { tool } from 'ai'
import { z } from 'zod'

/**
 * Looks up the current temperature, humidity and weather for the given location.
 */
export const getWeatherTool = tool({
  description: 'Looks up the current temperature, humidity and weather for the given location.',
  parameters: z.object({
    location: z.string().describe('The city or district to look up the weather for (e.g. Istanbul, TR).'),
    unit: z.enum(['celsius', 'fahrenheit']).describe('The unit used to display the temperature.').optional(),
  }),
  execute: async ({ location, unit }) => {
    // TODO: Run your real API call or business logic here
    return {
      success: true,
      timestamp: new Date().toISOString(),
      data: { location, unit },
    }
  },
})

AI Function Calling & Tool Use: The Bridge That Connects Agents to the World

Traditional chatbots only generate text, while modern AI agents perform real-world actions such as querying databases, calling APIs, sending emails or running financial calculations through the Function Calling (Tool Use) mechanism.

For LLMs to call tools at the right time with the right parameters, the function schemas (JSON Schema) and parameter descriptions must be defined without errors. This tool lets you generate the Vercel AI SDK 4.x (Zod), OpenAI Function Calling, Anthropic Claude Tools and TypeScript Interface formats from a single visual form in seconds.

Tips for Defining Effective Tools and Parameters

🎯 Description Quality

The AI model reads your parameter descriptions, not your code. Stating the parameter's expected format, unit (e.g. Celsius vs Fahrenheit) or a sample input clearly inside .describe() prevents 90% of hallucinations.

🛡️ Enums and Strict Type Limits

When there is a fixed set of values instead of open-ended strings, always use an enum. By stopping the model from producing arbitrary text, you get deterministic and safe API calls.

⚡ The Vercel AI SDK 4.x Zod Advantage

The Vercel AI SDK automatically converts a Zod schema into OpenAI/Claude-compatible JSON Schema. It gives you type safety and runs async business logic directly in the execute block.

Frequently Asked Questions (FAQ)

No, the LLM does not run code directly. To answer the user, the model tells you in JSON which tool it wants to call with which arguments (a tool call). Your server runs that call and sends the result back to the model, which then produces the final answer.

Vercel AI SDK 4.x (Zod), OpenAI Function Calling (JSON Schema), Anthropic Claude Tools and TypeScript Interface formats are supported.

OpenAI uses a { type: "function", function: { ... } } wrapper, while Anthropic Claude uses the { name, description, input_schema } structure directly. The tool generates both formats in full.

Pass the generated tool definition into streamText({ model, tools: { myTool }, maxSteps: 5 }). With maxSteps, the model can run the function automatically, read its output and reply to the user.