CodeGraph

An MCP server that uses AST parsing to map TypeScript/JavaScript codebase structure and expose codebase intelligence for AI assistants.

Features

• AST-based analysis - Uses ts-morph (TypeScript compiler API) to parse code without running it
• Dependency graphs - Builds graph representations of how files, functions, and classes relate
• Impact ranking - Identifies the most central/critical files using graph centrality metrics
• Call chain tracing - Finds dependency paths between any two symbols
• Mermaid diagrams - Generates visual dependency diagrams

MCP Tools

Tool	Description
scan_codebase	Scan a directory and return a summary of all files, functions, classes, and relationships
find_function	Search for a function or class by name, returns location, signature, callers, and callees
analyze_dependencies	Returns the full dependency graph or a subgraph for a specific file (JSON or Mermaid)
rank_impact	Ranks files by centrality (in-degree, out-degree, betweenness) to identify critical modules
trace_call_chain	Traces the call chain / dependency path from one function or file to another
analyze_complexity	Analyzes code complexity metrics for each file (cyclomatic, cognitive complexity, nesting depth)

MCP Resources

Resource	Description
codebase://summary	Cached summary of the most recently scanned codebase
codebase://graph/{format}	Dependency graph in json or mermaid format

Setup

Install

npm install
npm run build

Use with Claude Desktop (or any MCP client)

Add this to your MCP client configuration:

{
  "mcpServers": {
    "CodeGraph": {
      "command": "npx",
      "args": ["tsx", "/absolute/path/to/CodeGraph/src/index.ts"]
    }
  }
}

Or after building:

{
  "mcpServers": {
    "CodeGraph": {
      "command": "node",
      "args": ["/absolute/path/to/CodeGraph/dist/index.js"]
    }
  }
}

Usage

Once connected, AI assistants can:

1. Scan your codebase: "Scan my codebase at ./src"
2. Find symbols: "Where is the authenticate function defined and who calls it?"
3. Analyze dependencies: "Show me the dependency graph as mermaid"
4. Rank impact: "What's the most central file in my project?"
5. Trace paths: "Trace the call chain from handleRequest to saveToDatabase"

Architecture

src/
├── index.ts              # MCP server entry point (stdio transport)
├── parser/
│   ├── types.ts          # FileInfo, FunctionInfo, ClassInfo, etc.
│   ├── FileAnalyzer.ts   # ts-morph single-file analysis
│   └── ProjectParser.ts  # Directory scanning with caching
├── graph/
│   ├── types.ts          # GraphNode, GraphEdge, RankedFile, etc.
│   ├── GraphBuilder.ts   # Converts ParseResult to graphology Graph
│   └── GraphAnalyzer.ts  # Centrality, path finding, cycle detection
└── mcp/
    ├── cache.ts          # Shared analyzer cache
    ├── tools.ts          # 5 MCP tool definitions
    └── resources.ts      # MCP resource definitions

Tech Stack

• ts-morph - TypeScript AST parsing (wraps the TypeScript compiler API)
• graphology - Graph data structure for dependency mapping
• graphology-metrics - Centrality algorithms (betweenness, degree, etc.)
• graphology-shortest-path - Bidirectional shortest path finding
• @modelcontextprotocol/server - MCP server SDK with stdio transport
• zod - Schema validation for tool inputs

Ignored Directories

By default, these directories are excluded from scanning:

• node_modules, dist, build, .git, coverage
• .next, .nuxt, .svelte-kit, __tests__, .cache

License

MIT