🧠 CodeBrain MCP Server

Semantic code search powered by AI embeddings and vector similarity.

Integrate intelligent code search directly into Claude Desktop and Cursor through the Model Context Protocol (MCP).

🎯 What It Does

CodeBrain indexes your codebase using AST-based splitting and AI embeddings, enabling:

• Semantic search - Find code by meaning, not just keywords
• Smart chunking - AST-aware code splitting (respects functions, classes, etc.)
• Fast retrieval - Vector similarity search with pgvector
• Multi-project - Index and search across multiple codebases

🚀 Quick Start

1. Prerequisites

# Docker running (for PostgreSQL + pgvector)
docker ps | grep codebrain

# Node.js 20+
node --version

# Dependencies installed
cd /Users/conorandrle/Documents/Coding/CodeBrainMCP/CodeBrain
pnpm install

2. Setup Database

# Start PostgreSQL with pgvector (if not running)
docker run -d \
  --name codebrain \
  -p 5484:5432 \
  -e POSTGRES_PASSWORD=postgres \
  -e POSTGRES_DB=codebrain \
  pgvector/pgvector:pg15

# Setup database schema
pnpm db:setup
pnpm db:migrate

3. Configure Environment

Edit .env:

GEMINI_API_KEY=your_api_key_here
DATABASE_URL=postgresql://postgres:postgres@localhost:5484/codebrain?schema=cbmcp

4. Test the Server

# Run tests
pnpm test

# Should show: ✅ 28 tests passed

# Test MCP server starts
npx tsx src/index.ts
# Should output: 🚀 CodeBrain MCP Server started (stdio mode)
# Press Ctrl+C to stop

🔌 Connect to Cursor/Claude

For Cursor

1. Open Cursor Settings → MCP Servers
2. Add server named codebrain
3. Copy this config:

{
  "command": "npx",
  "args": [
    "-y",
    "tsx",
    "/Users/conorandrle/Documents/Coding/CodeBrainMCP/CodeBrain/src/index.ts"
  ],
  "env": {
    "GEMINI_API_KEY": "your_key_here",
    "DATABASE_URL": "postgresql://postgres:postgres@localhost:5484/codebrain?schema=cbmcp"
  }
}

4. Restart Cursor
5. Verify - Check MCP panel shows "codebrain" connected

📖 Detailed guide: See CURSOR_SETUP.md

For Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "codebrain": {
      "command": "npx",
      "args": [
        "-y",
        "tsx",
        "/Users/conorandrle/Documents/Coding/CodeBrainMCP/CodeBrain/src/index.ts"
      ],
      "env": {
        "GEMINI_API_KEY": "your_key_here",
        "DATABASE_URL": "postgresql://postgres:postgres@localhost:5484/codebrain?schema=cbmcp"
      }
    }
  }
}

Restart Claude Desktop.

🛠️ Available MCP Tools

1. index_codebase

Index a codebase for semantic search.

Parameters:

{
  projectName: string;   // Unique project identifier
  rootPath: string;      // Absolute path to code
  force?: boolean;       // Re-index existing files
}

Example:

"Index my React project at /Users/me/projects/my-app with name 'my-app'"

2. semantic_search

Search code semantically across indexed projects.

Parameters:

{
  query: string;         // What to search for
  projectName?: string;  // Filter by project
  topK?: number;        // Number of results (default: 5)
  threshold?: number;   // Similarity threshold (default: 0.5)
}

Example:

"Find authentication logic in my-app"

3. list_projects

List all indexed projects.

Parameters: None

Example:

"Show me all indexed projects"

4. get_project_stats

Get statistics for a project.

Parameters:

{
  projectName: string;   // Project to query
}

Example:

"Show me stats for the my-app project"

📊 Architecture

┌─────────────────────────────────────────────┐
│         Cursor / Claude Desktop             │
│              (MCP Client)                   │
└─────────────────┬───────────────────────────┘
                  │ MCP Protocol (stdio)
                  │
┌─────────────────▼───────────────────────────┐
│         CodeBrain MCP Server                │
│  ┌─────────────────────────────────────┐   │
│  │  AST Code Splitter                  │   │
│  │  - JavaScript/TypeScript            │   │
│  │  - Python, Go, Rust, Java, C++      │   │
│  └──────────────┬──────────────────────┘   │
│                 │                           │
│  ┌──────────────▼──────────────────────┐   │
│  │  Gemini Embeddings                  │   │
│  │  - 768-dimensional vectors          │   │
│  │  - Semantic descriptions            │   │
│  └──────────────┬──────────────────────┘   │
│                 │                           │
│  ┌──────────────▼──────────────────────┐   │
│  │  Vector Search                      │   │
│  │  - Cosine similarity                │   │
│  │  - Threshold filtering              │   │
│  └──────────────┬──────────────────────┘   │
└─────────────────┼───────────────────────────┘
                  │
┌─────────────────▼───────────────────────────┐
│      PostgreSQL + pgvector                  │
│  ┌──────────────────────────────────────┐  │
│  │  Projects → Files → Chunks → Embeds │  │
│  │  Normalized relational schema        │  │
│  └──────────────────────────────────────┘  │
└─────────────────────────────────────────────┘

🧪 Testing

# Run all tests
pnpm test

# Watch mode
pnpm test:watch

# Individual test suites
pnpm test:splitter     # AST code splitter
pnpm test:indexing     # Indexing workflow
pnpm test:search       # Semantic search
pnpm test:embedding    # Embedding generation

# Integration test (end-to-end)
pnpm test:integration

🌐 Graph Viewer (React)

Visualise the code graph in the browser with the React/Vite viewer.

# Start the Graph API server (serves graph JSON on http://localhost:4000)
pnpm graph:server

# In a separate terminal, install and run the viewer UI
cd apps/graph-viewer
pnpm install
pnpm dev

# Open the browser UI → http://localhost:5173

Override the API target with VITE_GRAPH_API_URL (inside apps/graph-viewer/.env) if the server runs elsewhere.

📁 Project Structure

CodeBrain/
├── src/
│   ├── index.ts              # MCP server entry point
│   ├── core/
│   │   ├── indexing.ts       # Indexing orchestration
│   │   ├── search.ts         # Semantic search
│   │   ├── splitter.ts       # AST-based code splitting
│   │   └── embedding/
│   │       ├── base-embedding.ts      # Embedding interface
│   │       └── gemini-embedding.ts    # Gemini implementation
│   └── test/
│       ├── *.test.ts         # Unit tests
│       └── utils.ts          # Test utilities
├── db/
│   ├── index.ts              # Prisma client
│   ├── setup.ts              # Database setup script
│   └── vector-indexes.ts     # Vector index management
├── prisma/
│   └── schema.prisma         # Database schema
├── .env                      # Environment variables
├── mcp-config.json          # MCP configuration template
├── CURSOR_SETUP.md          # Cursor integration guide
└── README.md                # This file

🗃️ Database Schema

Project (1) ─┐
             ├─> File (N) ─┐
                           ├─> Chunk (N) ─┐
                                          ├─> Embedding (N)

• Project: Root container (name, rootPath)
• File: Individual source files (path, language, hash)
• Chunk: Code segments (text, lines, AST metadata)
• Embedding: Vector representations (768-dim, model, similarity search)

🔧 Development

Scripts

pnpm dev              # Start with auto-reload
pnpm start            # Start server
pnpm build            # Compile TypeScript

pnpm db:setup         # Setup database + pgvector
pnpm db:migrate       # Run migrations
pnpm db:generate      # Generate Prisma client
pnpm db:studio        # Open Prisma Studio

Environment Variables

# Required
GEMINI_API_KEY=your_gemini_api_key
DATABASE_URL=postgresql://user:pass@host:port/db?schema=cbmcp

# Optional
NODE_ENV=development

🐛 Troubleshooting

MCP Connection Issues

Problem: Server won't connect in Cursor

Solutions:

1. Test manually: npx tsx src/index.ts (should output startup message)
2. Check absolute path in config matches your directory
3. Verify environment variables in MCP config
4. Restart Cursor completely (Cmd+Q, then reopen)
5. Check MCP output panel for error logs

Database Issues

Problem: type "vector" does not exist

Solution:

pnpm db:setup  # This installs pgvector in cbmcp schema

Problem: Connection refused

Solution:

docker ps | grep codebrain  # Verify container running
docker start codebrain      # Start if stopped

Embedding Issues

Problem: GEMINI_API_KEY is required

Solution: Add API key to .env and MCP config

Performance Issues

Problem: Indexing is slow

Solutions:

• Embeddings are cached - subsequent runs are faster
• Adjust batch size in indexing.ts if needed
• Consider excluding large directories (node_modules, etc.)

📈 Performance

• Indexing: ~2-5 seconds per file (first time, includes embedding generation)
• Re-indexing: ~100ms per file (if unchanged, uses hash comparison)
• Search: ~500ms per query (includes embedding + vector search)
• Storage: ~10KB per code chunk (text + embedding + metadata)

🔐 Security

• API keys stored in environment variables (not in code)
• Database credentials configurable
• MCP runs locally (no external API calls except Gemini)
• Vector embeddings don't leave your machine

📝 License

MIT

🤝 Contributing

This is a personal project, but feel free to fork and adapt for your needs!

🎓 Learn More

• Model Context Protocol
• pgvector
• Gemini Embeddings
• Tree-sitter

✅ Status

• ✅ Database setup and migrations
• ✅ AST-based code splitting
• ✅ Gemini embedding integration
• ✅ Vector similarity search
• ✅ MCP server implementation
• ✅ Comprehensive test suite (28 tests)
• ✅ Multi-project support
• ✅ Cursor/Claude integration ready

Ready for production use! 🚀