CodeInteliMCP - Intelligent Code Analysis Server

An advanced MCP (Model Context Protocol) server that combines Tree-sitter parsing with graph databases and vector search for lightning-fast code intelligence across multiple repositories.

🚀 Features

• Multi-Repository Support: Manage and analyze multiple codebases simultaneously
• Instant Usage Search: Find where functions/classes are used in milliseconds
• Semantic Code Search: Find similar code patterns using vector embeddings
• Dependency Analysis: Visualize and query code dependencies
• Impact Analysis: See what breaks when you change something
• Auto-indexing: Smart incremental updates as you code
• Multi-language Support: Python, JavaScript, TypeScript, Go, Rust, and more

🏗 Architecture

┌─────────────────┐     ┌──────────────────┐     ┌─────────────────┐
│  Claude Code    │────▶│   MCP Server     │────▶│   Tree-sitter   │
└─────────────────┘     └──────────────────┘     └─────────────────┘
                               │                           │
                               ▼                           ▼
                        ┌──────────────┐          ┌────────────────┐
                        │  DuckDB      │          │  ChromaDB      │
                        │  (Graph)     │          │  (Vectors)     │
                        └──────────────┘          └────────────────┘

🎯 Core Capabilities

1. Graph Queries (via DuckDB)

• Function call relationships
• Import dependencies
• Class inheritance trees
• Variable usage tracking

2. Vector Search (via ChromaDB)

• Semantic code similarity
• Natural language code search
• Pattern detection
• Code duplication finding

3. Real-time Analysis (via Tree-sitter)

• On-demand parsing
• Syntax validation
• Code structure analysis
• AST manipulation

🚦 Quick Start

Automated Setup (Recommended)

# Clone the repository
git clone https://github.com/rahulvgmail/CodeInteliMCP.git
cd CodeInteliMCP

# Run the setup script
python setup_for_claude.py

The setup script will:

• Install all dependencies
• Configure Claude Desktop/Code
• Create data directories
• Set up environment variables

Manual Setup

# Install dependencies
pip install mcp duckdb chromadb sentence-transformers aiofiles
pip install tree-sitter tree-sitter-python tree-sitter-javascript

# Add to Claude Desktop config (~/.config/claude/claude_desktop_config.json):
{
  "mcpServers": {
    "code-intelligence": {
      "command": "python",
      "args": ["/path/to/CodeInteliMCP/code_intelligence_mcp/server_minimal.py"],
      "env": {
        "CODE_INTEL_PROJECT_ROOT": "/path/to/your/project"
      }
    }
  }
}

🛠️ Available Tools

This MCP server provides the following tools:

Repository Management

• add_repository - Add a new repository to track

  • name (string, required): Repository name
  • path (string, required): Path to repository
  • description (string, optional): Repository description
  • make_active (boolean, optional): Set as active repository

• list_repositories - List all tracked repositories

• set_active_repository - Set the active repository for operations

  • name (string, required): Repository name to make active

• index_repository - Index an entire repository

  • repository_name (string, required): Name of repository to index
  • include_patterns (string, optional): Comma-separated file patterns to include
  • exclude_patterns (string, optional): Comma-separated file patterns to exclude

Code Analysis

• index_file - Index a single file into the code graph

  • file_path (string, required): Path to the file to index

• find_symbol - Find a symbol by name

  • name (string, required): Symbol name to search for
  • file_path (string, optional): Limit search to specific file

• find_usages - Find all usages of a symbol

  • name (string, required): Symbol name to find usages for
  • file_path (string, optional): Limit search to specific file

• semantic_search - Search for code using semantic similarity

  • query (string, required): Natural language search query
  • limit (integer, optional): Maximum number of results (default: 10)
  • symbol_types (string, optional): Comma-separated list of symbol types to filter

• find_similar_code - Find code similar to a given symbol

  • symbol_name (string, required): Reference symbol name
  • file_path (string, optional): File containing the reference symbol

System Tools

• test_connection - Test that the server is working

• get_index_stats - Get current index statistics

• get_vector_stats - Get vector store statistics

📖 Usage Examples

Once configured, use these tools in Claude:

# Add a new repository
add_repository(name="my-app", path="/path/to/my-app", description="Main application")

# Index the repository
index_repository(repository_name="my-app", include_patterns="**/*.py,**/*.js")

# Search for authentication functions
semantic_search(query="authentication and login functions", limit=20)

# Find where a class is used
find_usages(name="UserService")

# Find similar implementations
find_similar_code(symbol_name="login_user")

📚 Documentation

• USER_GUIDE.md: Complete setup and usage guide
• ARCHITECTURE.md: Technical architecture details
• IMPLEMENTATION_PLAN.md: Development roadmap

📊 Performance Targets

• Initial indexing: < 1 minute for 100k LOC
• Incremental updates: < 1 second per file
• Usage queries: < 50ms
• Semantic search: < 200ms