mcp-chain-of-draft-prompt-tool

概述

MCP 链式草稿提示工具将标准提示转换为链式草稿格式，通过最多减少 92.4% 的令牌使用量来增强推理能力。该工具支持多种 LLM 提供商，包括 Anthropic Claude、OpenAI GPT 模型、Mistral AI 以及通过 Ollama 或自定义端点使用的本地模型。工具自动处理提示转换、LLM 处理和结果格式化，同时保持高精度。

试试问 AI

装完之后，这里有 5 个你可以让 AI 做的事：

• chain_of_draft_solve

  Solve a problem using Chain of Draft reasoning

  试试:Solve this complex problem using Chain of Draft reasoning试试:Apply CoD to answer: What causes climate change?
• math_solve

  Solve a math problem with Chain of Draft

  试试:Calculate 247 + 394 using Chain of Draft试试:Solve: What is the square root of 144?
• code_solve

  Solve a coding problem with Chain of Draft

  试试:Debug this Python code using Chain of Draft试试:Write a function to sort an array with CoD
• logic_solve

  Solve a logic problem with Chain of Draft

  试试:Solve this riddle using Chain of Draft reasoning试试:Apply CoD to: If all roses are flowers, and some flowers fade, do some roses fade?
• get_performance_stats

  Get performance statistics for Chain of Draft vs Chain of Thought

  试试:Show me performance stats between CoD and CoT试试:Compare token usage between reasoning methods
• get_token_reduction

  Get token reduction statistics when using Chain of Draft

  试试:How many tokens did Chain of Draft save?试试:Show token reduction statistics
• analyze_problem_complexity

  Analyze the complexity of a given problem

  试试:Analyze the complexity of this math problem试试:How complex is this coding challenge?

可对比工具

安装

{
    "mcpServers": {
        "chain-of-draft-prompt-tool": {
            "command": "python3",
            "args": ["/absolute/path/to/cod/server.py"],
            "env": {
                "ANTHROPIC_API_KEY": "your_api_key_here"
            }
        }
    }
}

FAQ

什么是链式草稿（CoD）推理？

链式草稿是一种推理方法，使用非常简洁的步骤（通常不超过 5 个单词）来保持准确性，同时显著减少与传统链式思维（CoT）相比的令牌使用量。

支持哪些 LLM 提供商？

该工具支持 Anthropic Claude、OpenAI GPT 模型、Mistral AI，以及通过 Ollama 或兼容 OpenAI API 的自定义端点使用的本地模型。

mcp-chain-of-draft-prompt-tool 对比