Unverified50% confidenceFactExact time
在Cursor使用场景中，发送给模型的上下文包括编辑器自动附加的代码上下文、文件结构信息和项目索引数据，隐含Token消耗往往是用户可见输入的数倍甚至数十倍
1
Sources
50%
Confidence
Long-term
Relevance
8/16/2026
First Seen
Sources
Related Claims
UnverifiedIn programming assistant scenarios, system prompts and code context typically consume a large number of tokens, resulting in an extremely high cache hit rate.75% similarUnverified标注操作会生成带坐标信息的结构化 Prompt，将用户空间指向意图转化为模型可精确定位的上下文70% similarUnverifiedIn agent scenarios, due to multiple rounds of tool calls, each call must carry the previous conversation history and system prompt, causing input tokens to grow cumulatively, and a code refactoring task may consume tens of thousands or even hundreds of thousands of tokens.70% similarUnverifiedContinue 在生成代码时会自动收集工作区上下文信息，包括已打开文件内容、项目目录结构、语言配置和光标附近代码片段注入 Prompt70% similarUnverified更大的上下文窗口意味着可以上传更完整的信息，模型看到的信息越完整推理质量越高69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/757549API
curl https://kongchang.com/api/v1/knowledge/claims/757549MCP
get_claim(id=757549)