Verified65% confidenceFactExact time
The tool-calling capability of a large model itself actually only outputs a structured piece of JSON data describing which function to call and what parameters to pass; the model itself has no ability to execute code.
3
Sources
65%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedBy forcing the model to answer using a single tool, you can obtain JSON output that conforms to a schema rather than free-form text that needs to be parsed manually.76% similarVerifiedUnder the tool-calling paradigm, the LLM outputs a structured function-call intent (in JSON format), which is actually executed by an external system that returns the result to the model.76% similarUnverifiedJSON Schema annotation keywords such as title, description, and examples do not affect validation logic, but large models rely on these annotations to understand parameter meanings76% similarUnverifiedLLM本身并不直接执行工具，而是生成一个JSON字符串指定要调用的工具名和参数，由Agent执行逻辑解析并实际调用对应函数76% similarVerifiedTools are essentially function wrappers for external capabilities, paired with JSON Schema descriptions that enable the model to understand a tool's functionality and parameter format.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/897085API
curl https://kongchang.com/api/v1/knowledge/claims/897085MCP
get_claim(id=897085)