How to Use the LLM Function Calling & Structured Outputs Builder
-
1
Enter your tool name and descriptive purpose (used by the LLM to decide when to call the tool).
-
2
Add parameters: specify field names, types (string, number, boolean, array), descriptions, and required status.
-
3
Switch output format between OpenAI Tools, Anthropic Claude Tools, or Python Pydantic classes.
-
4
Copy the ready-to-use specification directly into your Python or TypeScript application.
Practical Example & Use Case
Building a strict schema for an e-commerce order lookup tool with required `order_id` (string) and optional `include_tracking_history` (boolean) parameters.
Technical Notes & Limitations
Generates valid JSON Schema and Pydantic declarations. Works 100% offline in your browser.
Frequently Asked Questions
What is LLM function calling?
Function calling allows language models to connect to external APIs by outputting a structured JSON object containing verified arguments that match your schema, instead of unstructured free text.
What is the difference between OpenAI and Anthropic tool formats?
OpenAI nests the schema under `{ type: "function", function: { name, description, parameters } }`. Anthropic uses `{ name, description, input_schema }`. This tool exports both formats with 1 click.
Why are parameter descriptions so important?
LLMs rely directly on parameter descriptions to understand format expectations (e.g. "ISO-8601 date string YYYY-MM-DD") and argument requirements.