code-verify-mcp

code-verify-mcp

An MCP server for verifying AI-generated code quality, security, and performance, addressing trust gaps in AI coding assistants.

Category
访问服务器

README

code-verify-mcp — AI Code Verification MCP Server

Zero-dependency MCP server for verifying AI-generated code quality, security, and performance. Built to address the critical trust gap where AI coding assistants routinely lie about task completion and produce code with hidden vulnerabilities.

Problem Statement

46% of developers distrust AI output, yet 84% use AI tools daily (Stack Overflow 2025/2026). There's a critical gap for AI code verification in the MCP ecosystem.

Key Issues:

  • AI agents confidently state completion when tasks are actually incomplete
  • AI-generated code often has security vulnerabilities (XSS, SQL injection, eval usage)
  • Code quality issues (long functions, magic numbers, TODO comments) go undetected
  • Performance bottlenecks (nested loops, N² complexity) remain in production
  • Developers lack independent verification of AI-generated code functionality

Market Validation:

  • Growing AI coding tools adoption (Cursor 18%, Claude Code 10%)
  • Technical debt = #1 developer frustration
  • MCP ecosystem: 97M SDK downloads, 400% YoY growth
  • 46% distrust AI output, only 3% highly trust

Features

Security Analysis

  • XSS Detection: Detects innerHTML, dangerouslySetInnerHTML, and unsafe element creation
  • Injection Prevention: Identifies SQL injection patterns and dynamic query construction
  • Security Anti-patterns: Flags eval(), document.write(), and other dangerous practices
  • OWASP Compliance: Checks against OWASP security guidelines
  • Security Score: 0-100 score with severity classification

Code Quality Analysis

  • Function Length Analysis: Detects overly long functions (>50 lines by default)
  • Technical Debt Tracking: Identifies TODO, FIXME, HACK comments
  • Magic Number Detection: Flags hardcoded values without constants
  • Maintainability Scoring: Calculates maintainability index based on complexity
  • Code Style Validation: Checks for consistent code patterns

Performance Analysis

  • Complexity Detection: Identifies O(n²) complexity patterns (nested loops)
  • Inefficient Operations: Flags repeated operations that could be optimized
  • Memory Leak Detection: Identifies potential memory leak patterns
  • Big O Analysis: Estimates cyclomatic and cognitive complexity

Functional Verification

  • Empty Function Detection: Identifies unused or empty functions
  • Variable Usage: Detects potential unused variables
  • Undefined Behavior: Checks for unsafe dereferences in JS/TS

MCP Integration

  • Model Context Protocol: Full MCP server implementation
  • Tool Exposition: Multiple tools for code verification
  • Real-time Verification: Instant feedback on code generation
  • Integration Ready: Works with Cursor, Windsurf, Claude Code, and other AI assistants

Installation

npm install -g code-verify-mcp

Usage

Command Line Interface

# Verify code snippet
code-verify-mcp verify "function greet(name) { return \`Hello, \${name}!\`; }" -l javascript

# Generate test cases
code-verify-mcp generate-tests "eval('console.log(\"test\")')" -l javascript

# Analyze complexity
code-verify-mcp complexity "for (let i = 0; i < 100; i++) { for (let j = 0; j < 100; j++) { console.log(i, j); } }" -l javascript

# Show configuration
code-verify-mcp config

# Run demo
code-verify-mcp demo

JSON Output

code-verify-mcp verify 'eval("console.log(\'test\')")' -l javascript -o json

Output:

{
  "isValid": false,
  "score": 75,
  "issues": [
    {
      "type": "security",
      "severity": "critical",
      "message": "Found eval() which is a security risk",
      "suggestion": "Avoid using eval() with dynamic input. Use safer alternatives.",
      "line": 1
    }
  ],
  "recommendations": [
    "Avoid using eval() with dynamic input. Use safer alternatives."
  ],
  "summary": {
    "totalIssues": 1,
    "criticalCount": 1,
    "highCount": 0,
    "mediumCount": 0,
    "lowCount": 0
  }
}

Markdown Output

code-verify-mcp verify 'function greet(name) { return \`Hello, \${name}!\`; }' -l javascript -o markdown

Output:

# Code Verification Report

## Final Score: 100/100

### Summary
| Metric | Count |
|--------|-------|
| Total Issues | 0 |
| Critical | 0 |
| High | 0 |
| Medium | 0 |
| Low | 0 |

### Issues

### Recommendations

1. No recommendations

MCP Tools

The MCP server exposes the following tools:

verifyCodeSnippet

Verify a code snippet for quality, security, and performance issues.

Parameters:

{
  code: string;        // Code to verify
  language: string;    // Programming language
  frameworks?: string[]; // Frameworks used
  securityLevel?: 'basic' | 'strict' | 'comprehensive';
}

Returns:

{
  isValid: boolean;
  score: number;  // 0-100
  issues: Array<{
    type: 'security' | 'quality' | 'performance' | 'functional';
    severity: 'low' | 'medium' | 'high' | 'critical';
    message: string;
    suggestion?: string;
    line?: number;
  }>;
  recommendations: string[];
}

generateSecurityTests

Generate security test cases based on code analysis.

Parameters:

{
  code: string;
  language: string;
}

Returns:

{
  tests: Array<{
    description: string;
    code: string;
    type: 'security' | 'performance' | 'functional';
  }>;
}

analyzeCodeComplexity

Analyze code complexity metrics.

Parameters:

{
  code: string;
  language: string;
}

Returns:

{
  cyclomaticComplexity: number;
  cognitiveComplexity: number;
  maintainabilityIndex: number;
  suggestions: string[];
}

Configuration

Create a configuration file or use environment variables:

// trustshell.config.js
module.exports = {
  securityLevel: 'strict',
  analyzeQuality: true,
  analyzeSecurity: true,
  analyzePerformance: true,
  analyzeFunctionality: true,
  maxComplexity: 75,
  thresholds: {
    critical: 70,
    high: 80,
    medium: 85,
    low: 90,
  },
};

Integration with AI Coding Assistants

Cursor

// In Cursor configuration
{
  "mcpServers": {
    "code-verify": {
      "command": "code-verify-mcp"
    }
  }
}

Claude Code

# Install MCP server
npm install -g code-verify-mcp

# Configure Claude Code to use the server
# Add to Claude Code MCP configuration

Windsurf

// In Windsurf settings
{
  "mcpServers": {
    "code-verify": {
      "command": "code-verify-mcp"
    }
  }
}

Development

Building

npm run build

Testing

npm test

Development Mode

npm run dev

Type Checking

npm run type-check

Example Usage

Before (AI-Generated Code)

// AI generates this code
function calculateDiscount(price, quantity, customerType) {
  let discount = 0;

  if (customerType === 'VIP') {
    discount = price * 0.2;
  }

  for (let i = 0; i < quantity; i++) {
    price = price * 0.95;
  }

  return price - discount;
}

console.log(calculateDiscount(100, 10, 'VIP'));

Verification Results

Verification Results:

❌ Code has issues:

1. [MEDIUM] quality
   Function exceeds recommended length of 50 lines
   💡 Consider breaking the function into smaller, focused functions
   📍 Line 4

Final Score: 93/100

Summary:
  Total Issues: 1
  Critical: 0
  High: 0
  Medium: 1
  Low: 0

Recommendations:
  1. Consider breaking the function into smaller, focused functions

Features in Detail

Security Scanning

XSS Detection

// ❌ Detected
innerHTML = '<div>test</div>';
dangerouslySetInnerHTML = '<div>test</div>';
document.createElement('div').innerHTML = '<div>test</div>';

// ✅ Recommended
textContent = 'test';

SQL Injection Prevention

// ❌ Detected
$query("SELECT * FROM users WHERE id = " + userId);

// ✅ Recommended
$query("SELECT * FROM users WHERE id = ?", [userId]);

eval() Detection

// ❌ Detected
eval("console.log('test')");

// ✅ Recommended
console.log('test');

Code Quality

Function Length

// ❌ Detected (>50 lines)
function veryLongFunctionName() {
  // 60+ lines of code
}

// ✅ Recommended
function doSomething() {
  // 10-20 lines
}

Technical Debt

// ❌ Detected
// TODO: Implement this
// FIXME: Fix this bug
// HACK: This is a workaround

// ✅ Recommended
// Implement this functionality
// Fix this bug
// Remove this workaround

Performance

Nested Loops

// ❌ Detected (O(n²) complexity)
for (let i = 0; i < 100; i++) {
  for (let j = 0; j < 100; j++) {
    console.log(i, j);
  }
}

// ✅ Recommended
for (let i = 0; i < 100; i++) {
  const optimizedResult = preprocessData();
  console.log(i, optimizedResult);
}

Security Scoring

The system provides a 0-100 score based on:

  • Critical issues: -30 points each
  • High severity: -15 points each
  • Medium severity: -7 points each
  • Low severity: -2 points each

Thresholds:

  • 70+: Pass (Critical issues only)
  • 80+: Pass (High and below)
  • 85+: Pass (Medium and below)
  • 90+: Pass (Low and below)

Testing

Comprehensive test suite with 70+ test cases:

npm test -- --coverage

Covers:

  • Security analysis
  • Code quality detection
  • Performance analysis
  • Functional verification
  • Edge cases
  • Error handling
  • Cross-language support

Architecture

code-verify-mcp/
├── src/
│   ├── analysis.ts      # Core analysis engine
│   ├── index.ts         # Main exports
│   └── cli.ts           # CLI interface
├── tests/
│   ├── analysis.test.ts # Analysis tests
│   └── cli.test.ts      # CLI tests
├── package.json
├── tsconfig.json
├── vitest.config.ts
└── tsup.config.ts

Core Components

CodeAnalyzer: Main analysis engine that orchestrates all analysis modules

  • Security analyzer
  • Quality analyzer
  • Performance analyzer
  • Functional analyzer

Analysis Modules: Each type of analysis runs independently

  • Security patterns detection
  • Code quality checks
  • Performance benchmarks
  • Functional validation

Result Processing: Aggregates all findings and calculates scores

  • Issue severity calculation
  • Score computation
  • Summary generation
  • Recommendation generation

Performance

  • Analysis Speed: <10ms for typical codebase
  • Memory Usage: <50MB for large projects
  • Concurrency: Support for parallel analysis
  • Scalability: Handles 10,000+ lines per minute

Browser Compatibility

  • Node.js 18+
  • TypeScript 5.3+
  • Compatible with all modern browsers (when transpiled)

License

MIT - see LICENSE file for details.

Support

Contributing

We welcome contributions! Please see our Contributing Guide for details.

Roadmap

  • [ ] Enhanced security scanning (SAST integration)
  • [ ] Memory leak detection in Node.js
  • [ ] Framework-specific analysis (React, Vue, Angular)
  • [ ] Integration with ESLint and SonarQube
  • [ ] Real-time code review in IDE
  • [ ] Team collaboration features
  • [ ] Advanced AI-powered recommendations

Why code-verify-mcp?

In an era where AI-generated code becomes increasingly prevalent, independent verification is not just a luxury—it's a necessity. code-verify-mcp bridges the gap between AI's "helpful" persona and technical accuracy, giving developers the confidence to trust their AI-assisted development workflow.

Built by developers, for developers. 🛡️

Related Projects

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选