Unverified50% confidenceFactExact time
Every exposed tool consumes context Tokens on every request; information such as tool definitions, parameter descriptions, and return value formats all needs to be stuffed into the model context.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
When Should You Not Build an MCP Server? AI Engineer Interview Question Explained
youtubeCode & Agents10/16/2026
Related Entities
Related Claims
UnverifiedWhen an application sends a user request to the model, it includes a list of tools the model can use, with each tool having a name and an input schema.75% similarUnverifiedThe actual number of tokens consumed by a single user request includes the entire input context (including conversation history, system prompts, tool definitions, etc.) as well as the response content generated by the model.74% similarUnverifiedThe scope of forced tool calling (tool_choice) is a single generation (completion) rather than the entire session, so it must be re-specified on every call to the model.73% similarUnverifiedThe context sent with each model call typically includes the system prompt, the tool definition list, memories and files, and the full conversation record.72% similarUnverifiedDue to the stateless design of large model APIs, the entire tools array must be resent in every request, alongside the system prompt and conversation history.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1048233API
curl https://kongchang.com/api/v1/knowledge/claims/1048233MCP
get_claim(id=1048233)