deep-code-reasoning-mcp

概述

该 MCP 服务器创建了一个多模型工作流，其中 Claude Code 处理终端集成和多文件重构，而 Gemini 则利用其巨大的上下文窗口（1M tokens）和代码执行能力进行分布式系统调试和长链分析。它根据两个 AI 系统的特定优势，在模型之间提供智能路由。

试试问 AI

装完之后，这里有 5 个你可以让 AI 做的事：

• start_conversation{claude_context: {attempted_approaches: string[], partial_findings: any[], stuck_description: string, code_scope: {files: string[], entry_points?: CodeLocation[], service_names?: string[]}}, analysis_type: 'execution_trace' | 'cross_system'

  Initiates a conversational analysis session between Claude and Gemini.

  试试:Start a conversation to analyze performance issues in the UserService试试:Begin a cross-system impact analysis of these code changes
• continue_conversation{session_id: string, message: string, include_code_snippets?: boolean}

  Continues an active conversation with Claude's response or follow-up question.

  试试:Continue the conversation with a question about query optimization试试:Follow up with a request for more context about the race condition
• finalize_conversation{session_id: string, summary_format: 'detailed' | 'concise' | 'actionable'}

  Completes the conversation and generates structured analysis results.

  试试:Finalize the conversation and provide actionable recommendations试试:Complete the analysis with a detailed summary of findings
• get_conversation_status{session_id: string}

  Checks the status and progress of an ongoing conversation.

  试试:What's the status of our conversation session?试试:Check if the analysis is complete
• escalate_analysis{claude_context: {attempted_approaches: string[], partial_findings: any[], stuck_description: string, code_scope: {files: string[], entry_points?: CodeLocation[], service_names?: string[]}}, analysis_type: 'execution_trace' | 'cross_system'

  Main tool for handing off complex analysis from Claude Code to Gemini.

  试试:Escalate this distributed system analysis to Gemini试试:Hand off the performance regression investigation
• trace_execution_path{entry_point: {file: string, line: number, function_name?: string}, max_depth?: number, include_data_flow?: boolean}

  Deep execution analysis with Gemini's semantic understanding.

  试试:Trace the execution path of this function试试:Analyze the data flow starting from this line of code
• cross_system_impact{change_scope: {files: string[], service_names?: string[]}, impact_types?: ('breaking' | 'performance' | 'behavioral')[]}

  Analyze impacts across service boundaries.

  试试:What's the cross-system impact of changing this API?试试:Analyze how these changes might affect other services
• performance_bottleneck{code_path: {entry_point: {file: string, line: number, function_name?: string}, suspected_issues?: string[]}, profile_depth?: 1-5}

  Deep performance analysis beyond simple profiling.

  试试:Identify performance bottlenecks in this function试试:Analyze the slow database queries in the user service
• hypothesis_test{hypothesis: string, code_scope: {files: string[], entry_points?: CodeLocation[]}, test_approach: string}

  Test specific theories about code behavior.

  试试:Test the hypothesis that the memory leak is caused by this function试试:Validate the theory that race conditions occur under high load

可对比工具

安装

git clone https://github.com/Haasonsaas/deep-code-reasoning-mcp.git
cd deep-code-reasoning-mcp

npm install

cp .env.example .env
# 编辑 .env 文件并添加您的 GEMINI_API_KEY

npm run build

{
  "mcpServers": {
    "deep-code-reasoning": {
      "command": "node",
      "args": ["/path/to/deep-code-reasoning-mcp/dist/index.js"],
      "env": {
        "GEMINI_API_KEY": "your-gemini-api-key"
      }
    }
  }
}

FAQ

这个服务器与其他 AI 代码分析工具有什么不同？

它结合了两个 AI 模型的优势 - Claude Code 用于本地上下文操作和 CLI 原生工作流，以及 Gemini 凭借其 1M token 上下文窗口进行大规模代码库分析和分布式系统调试。

使用此 MCP 服务器需要什么先决条件？

您需要 Node.js 18 或更高版本、具有 Gemini API 访问权限的 Google Cloud 账户，以及来自 Google AI Studio 的 Gemini API 密钥。

deep-code-reasoning-mcp 对比