Defining Schemas & Forcing Execution — Advanced Tool Use

[ ABORT TO HUD ]
SEQ. 1
SEQ. 2

Defining Schemas & Forcing Execution

🔧 Advanced Tool Use⏱ 20 min⭐ 500 BASE XP⌨ HANDS-ON LAB

Building the Tool Socket

Claude interacts with your code via Tools (Function Calling), defined using standard JSON Schema (optionally setting strict: true for guaranteed constrained-decoding parameter conformance). Clear descriptions in input_schema act as direct instructions guiding when and how Claude invokes each tool.

The tool_choice Parameter

By default (auto), Claude decides when to use a tool. For deterministic pipelines, you can override this logic:

  • auto: Model decides probabilistic selection.
  • any: Forces Claude to use at least one tool from your list.
  • tool: Forces Claude to use a specific tool name immediately.
  • none: Prevents Claude from invoking any tools on that turn.
// Forcing a specific tool
"tool_choice": {"type": "tool", "name": "get_weather"}
⌨ HANDS-ON LABForce Tool Call with tool_choice
⭐ +200 XP

Define a lookup_weather tool schema with JSON Schema input parameters, and force Claude to invoke it using tool_choice: {"type": "tool", "name": "lookup_weather"}.

1POST to /v1/messages with a tools array and tool_choice set to force lookup_weather.
lab-sandbox — simulated environment
INFINITY LAB SANDBOX v2.6 — simulated shell
Type the command for the current objective. Helpers: "hint", "solution", "clear".
$
OBJECTIVE 1 / 1 — type "hint" if stuck
SYNAPSE VERIFICATION
QUERY 1 // 2
How does the 'input_schema' assist the model's reasoning?
It provides the executable code
It mathematically binds Claude's weights to output types that map to your code's requirements
It displays a UI block
It parses TypeScript