Giving AI Hands
An LLM in a box is smart but helpless. To be useful, it needs tools. Tool use (or function calling) allows models to interact with the outside world—sending emails, querying databases, or updating CRM records.
Defining Tools
The art of tool use lies in the definition. A good tool definition includes:
- •Clear Name: Unambiguous identification of what the tool does.
- •Detailed Description: Helping the model understand when and why to use it.
- •Typed Schema: Strict JSON schemas for arguments to ensure the model generates valid inputs.
The Feedback Loop
The most critical part of tool use is handling the output. When a tool executes, the result (or error) must be fed back to the model so it can reason about the next step. This "thought-action-observation" loop is the heartbeat of an agentic system.