Tarun097/code-search-mcp

Code Search MCP Server

A semantic code search MCP server that indexes code repositories using embeddings and provides intelligent search capabilities for Amazon Q CLI.

Features

• Semantic Search: Find code by meaning, not just keywords
• Keyword Search: Traditional text-based search
• Multi-Repository Support: Index and search across multiple repositories
• Chunk-based Indexing: Intelligent code chunking for better context
• File Type Awareness: Supports 20+ programming languages
• Real-time Integration: Works seamlessly with Amazon Q CLI

Architecture

Code Repo → Indexer → ChromaDB → MCP Server → Amazon Q CLI

Quick Start

1. Setup Environment

# Clone and setup
git clone <repository-url>
cd code-search-mcp

# Run quick setup
./quick_start.sh

2. Index Your First Repository

# Activate environment
source venv/bin/activate

# Index a repository
python indexer.py --repo-path /path/to/your/repo --repo-name my-project

# Verify indexing
python indexer.py --list

3. Start MCP Server

If this is your first time or after cleanup:

# Run quick setup to create virtual environment and install dependencies
./quick_start.sh

To start the server:

# Start the server (recommended)
./start_server_simple.sh

# Alternative startup script
./start_server.sh

Manual startup (if needed):

# Activate environment and start manually
source venv/bin/activate
python mcp_server.py

4. Configure Amazon Q CLI

# Add MCP server to Q CLI
qchat mcp add --name code-search --command /path/to/code-search-mcp/venv/bin/python --args /path/to/code-search-mcp/mcp_server.py

# Verify configuration
qchat mcp list

Repository Management

Index a New Repository

python indexer.py --repo-path /path/to/repo --repo-name repo-name

Example:

python indexer.py --repo-path /home/user/projects/my-app --repo-name my-app

List Indexed Repositories

python indexer.py --list

View Repository Statistics

# All repositories
python indexer.py --stats

# Specific repository
python indexer.py --stats --filter-repo my-app

Test Search Functionality

# Test semantic search
python indexer.py --search "authentication logic"

# Test search in specific repository
python indexer.py --search "database connection" --filter-repo my-app

Supported File Types

The indexer automatically processes these file types:

Programming Languages:

• Python (.py)
• JavaScript/TypeScript (.js, .ts)
• Java (.java)
• C# (.cs)
• C/C++ (.c, .cpp, .h)
• Go (.go)
• Rust (.rs)
• PHP (.php)
• Ruby (.rb)
• Swift (.swift)
• Kotlin (.kt)
• Scala (.scala)

Data & Config:

• JSON (.json)
• YAML (.yaml, .yml)
• XML (.xml)
• SQL (.sql)

Documentation:

• Markdown (.md)
• HTML (.html)
• CSS (.css)
• Text (.txt)

Usage with Amazon Q CLI

Once configured, you can ask Amazon Q CLI questions like:

Code Discovery

• "Find authentication logic in the my-app repository"
• "Show me database connection code"
• "Where is the user validation implemented?"
• "Find functions that handle file uploads"

API Documentation

• "What does the GetUserById API do?"
• "Show me all REST endpoints in the project"
• "Find error handling patterns"

Architecture Understanding

• "How is the application structured?"
• "Show me the service layer implementation"
• "Find dependency injection configuration"

MCP Tools Provided

Configuration Files

• q-cli-config.json: Amazon Q CLI MCP configuration
• chroma_db/repo_metadata.json: Repository metadata
• requirements.txt: Python dependencies

Troubleshooting

Starting After Git Clone or Cleanup

If you've just cloned the repository or cleaned up the environment:

1. Run quick setup:

   ./quick_start.sh
   

2. Index your repositories:

   source venv/bin/activate
   python indexer.py --repo-path /path/to/your/repo --repo-name repo-name
   

3. Start the MCP server:

   ./start_server_simple.sh
   

4. Configure Q CLI (if not done before):

   qchat mcp add --name code-search --command $(pwd)/venv/bin/python --args $(pwd)/mcp_server.py
   

MCP Server Not Loading

1. Check if the server is running: ps aux | grep mcp_server
2. Verify Q CLI configuration: qchat mcp list
3. Check server logs for errors

No Search Results

1. Verify repository is indexed: python indexer.py --list
2. Check repository path is correct
3. Ensure files are in supported formats

Performance Issues

1. Large repositories may take time to index
2. Consider indexing only specific directories
3. Monitor memory usage during indexing

Development

Project Structure

code-search-mcp/
├── indexer.py              # Repository indexing logic
├── mcp_server.py           # MCP server implementation
├── code_indexer.py         # Core indexing classes
├── start_server.sh         # Server startup script
├── quick_start.sh          # Setup script
├── requirements.txt        # Dependencies
├── chroma_db/             # Vector database
└── venv/                  # Python virtual environment

Adding New File Types

Edit code_indexer.py and add new extensions to the SUPPORTED_EXTENSIONS list.

Customizing Chunk Size

Modify the chunk_size parameter in CodeIndexer class initialization.

License

This project is owned by Saviynt, Inc. All rights reserved.

Quick Reference

Essential Commands

# Setup (first time or after cleanup)
./quick_start.sh

# Index a repository
source venv/bin/activate
python indexer.py --repo-path /path/to/repo --repo-name repo-name

# List indexed repositories
python indexer.py --list

# Start MCP server
./start_server_simple.sh

# Configure Q CLI
qchat mcp add --name code-search --command $(pwd)/venv/bin/python --args $(pwd)/mcp_server.py

# Check Q CLI configuration
qchat mcp list

File Structure After Setup

code-search-mcp/
├── venv/                   # Virtual environment (created by quick_start.sh)
├── chroma_db/             # Vector database (created when indexing)
├── mcp_server.py          # Main MCP server
├── indexer.py             # Repository indexer
├── start_server_simple.sh # Server startup script
└── README.md              # This file

Support

For issues and questions:

1. Check the troubleshooting section
2. Review MCP server logs
3. Verify Amazon Q CLI configuration
4. Test indexer functionality independently