A tool schema tells an AI system what an available function does and what arguments it accepts. A schema may define required fields, allowed values, descriptions and data types. Clear schemas reduce ambiguous calls and make it easier for the application to validate requests before execution.
A schema for an order status tool might require an order ID string and describe it as the identifier printed on the receipt. That description helps the model fill it correctly. Keep tool names and parameter descriptions specific, limit the number of tools offered at once, and test what happens when required fields are missing. Schemas are also the input to AI tool orchestration, and they are central to function calling and to MCP servers.
A schema is not an authorization system. Check user permissions and enforce constraints in the tool implementation itself, even if the model produces valid arguments. Related: Tool Calling, Structured Output.