A Model Context Protocol (MCP) server that analyzes codebases to generate dependency graphs and architectural insights. This server helps understand code structure, dependencies, and architectural patterns across multiple programming languages.

• Multi-Language Support: Analyzes dependencies in TypeScript, JavaScript, C#, Python, and more
• Dependency Graph Generation: Creates detailed dependency graphs in JSON or DOT format
• Architectural Analysis: Infers architectural layers and validates against rules
• File Metadata: Extracts imports, exports, and other metadata from source files
• Scoring System: Evaluates codebase against architectural rules and patterns

1. Clone the repository
2. Install dependencies:

npm install

3. Build the project:

npm run build

Add to your MCP settings file (usually located at ~/.config/cline/mcp_settings.json or equivalent):

json { mcpServers: { \DependencyMCP: { \command: \node, \args: [\path/to/dependency-mcp/dist/index.js], \env: { \MAX_LINES_TO_READ: \1000, \CACHE_DIR: \path/to/dependency-mcp/.dependency-cache, \CACHE_TTL: \3600000 } } }

Environment Variables:

• MAX_LINES_TO_READ: Maximum number of lines to read from each file (default: 1000)
• CACHE_DIR: Directory to store dependency cache files (default: .dependency-cache)
• CACHE_TTL: Cache time-to-live in milliseconds (default: 1 hour = 3600000)

Analyzes dependencies in a codebase and generates a dependency graph.

const result = await client.callTool("DependencyMCP", "analyze_dependencies", {
  path: "/path/to/project",
  excludePatterns: ["node_modules", "dist"], // optional
  maxDepth: 10, // optional
  fileTypes: [".ts", ".js", ".cs"] // optional
});

Gets the dependency graph for a codebase in JSON or DOT format.

const result = await client.callTool("DependencyMCP", "get_dependency_graph", {
  path: "/path/to/project",
  format: "dot" // or "json" (default)
});

Gets detailed metadata about a specific file.

const result = await client.callTool("DependencyMCP", "get_file_metadata", {
  path: "/path/to/file.ts"
});

Scores the codebase against architectural rules and patterns.

const result = await client.callTool("DependencyMCP", "get_architectural_score", {
  path: "/path/to/project",
  rules: [
    {
      pattern: "src/domain/**/*",
      allowed: ["src/domain/**/*"],
      forbidden: ["src/infrastructure/**/*"]
    }
  ]
});

{
  "src/index.ts": {
    "path": "src/index.ts",
    "imports": ["./utils", "./services/parser"],
    "exports": ["analyze", "generateGraph"],
    "namespaces": [],
    "architecturalLayer": "Infrastructure",
    "dependencies": ["src/utils.ts", "src/services/parser.ts"],
    "dependents": []
  }
}

{
  "score": 85,
  "violations": [
    "src/domain/user.ts -> src/infrastructure/database.ts violates architectural rules"
  ],
  "details": "Score starts at 100 and deducts 5 points per violation"
}

The server is built with TypeScript and uses:

• Zod for schema validation
• diff for file comparison
• minimatch for glob pattern matching

dependency-mcp/
├── src/
│   └── index.mts    # Main server implementation
├── package.json
├── tsconfig.json
└── README.md

To add support for a new programming language:

1. Add file extensions to the default fileTypes array
2. Implement language-specific regex patterns in parseFileImports and parseFileExports
3. Add any language-specific architectural patterns to inferArchitecturalLayer

MIT