AgenticRAG MCP Server
An intelligent codebase processing server that provides agentic RAG (Retrieval-Augmented Generation) capabilities through the Model Context Protocol (MCP).
Features
- Intelligent Code Indexing: Automatically chunks and embeds codebases for semantic search
- Agentic Retrieval: Self-critiquing retrieval loop that ensures comprehensive context
- Multi-Model Architecture: Uses GPT-4o for retrieval and Claude 3 for planning
- Live Updates: File system watching for automatic re-indexing
- Cost Control: Built-in telemetry and budget management
Quick Installation
1. Clone and Install
The install script will:
- Check Python version (3.8+ required)
- Create a virtual environment
- Install all dependencies
- Prompt for your API keys
- Create necessary directories
- Generate Claude configuration
2. Add to Claude
After installation, add AgenticRAG to Claude:
Windows (Claude Desktop):
- Open
%APPDATA%\Claude\claude_desktop_config.json
- Add the configuration from
claude_config_snippet.json
macOS/Linux (Claude Desktop):
- Open
~/.config/claude/claude_desktop_config.json
- Add the configuration from
claude_config_snippet.json
3. Restart Claude
Restart Claude to load the new MCP server.
Manual Installation
If you prefer to install manually:
Usage
Once installed, you can use these tools in Claude:
Initialize a Repository
Search Your Code
Get Repository Statistics
Example Conversation
Configuration
Required Environment Variables
Optional Configuration
Architecture
How It Works
- Indexing: The system chunks your code respecting language boundaries and creates embeddings
- Retrieval: When you search, an AI agent generates optimized queries and retrieves relevant chunks
- Self-Evaluation: The agent evaluates if it has enough context and can perform additional searches
- Compression: Results are intelligently summarized to provide clear, actionable answers
Troubleshooting
"No module named 'chromadb'"
Activate the virtual environment:
"OpenAI API key not found"
Make sure your .env
file contains:
"MCP server not found in Claude"
- Ensure you've added the configuration to Claude's config file
- Restart Claude Desktop completely
- Check the logs in
./logs/agenticrag.log
Search returns no results
Ensure you've indexed the repository first using the init_repo tool.
Development
Running Tests
Local Testing
Contributing
Contributions are welcome! Please:
- Fork the repository
- Create a feature branch (
git checkout -b feature/amazing-feature
) - Commit your changes (
git commit -m 'Add amazing feature'
) - Push to the branch (
git push origin feature/amazing-feature
) - Open a Pull Request
License
MIT License - see LICENSE file for details.
Acknowledgments
- Built for Claude Desktop using the Model Context Protocol
- Uses ChromaDB for vector storage
- Powered by OpenAI embeddings and LangGraph
If you prefer manual installation:
Usage
Once installed, you can use these tools in Claude:
Index a Repository
Search Code
Get Statistics
Example Conversation
Configuration
The server can be configured via environment variables in .env
:
Troubleshooting
Module Not Found
- Ensure virtual environment is activated:
source venv/bin/activate
- Check installation:
pip list | grep agenticrag
API Key Errors
- Verify keys in
.env
file - Ensure no extra spaces or quotes around keys
- Check key permissions for required models
Claude Can't Find Tools
- Verify configuration path is absolute, not relative
- Check Claude logs: Help → Show Logs
- Ensure MCP server section exists in config
Server Won't Start
- Check Python version:
python3 --version
(need 3.8+) - Verify Redis is running:
redis-cli ping
- Check port availability:
lsof -i:8000
Performance Issues
- Adjust
CHUNK_SIZE_TOKENS
for your codebase - Increase
EMBEDDING_BATCH_SIZE
for faster indexing - Monitor costs with
get_repo_stats
tool
Development
Running Tests
Code Formatting
Project Structure
Contributing
- Fork the repository
- Create a feature branch:
git checkout -b feature-name
- Make your changes and test
- Submit a pull request
License
MIT License - see LICENSE file for details
Support
- Issues: GitHub Issues
- Discussions: GitHub Discussions
- Documentation: Wiki
This server cannot be installed
hybrid server
The server is able to function both locally and remotely, depending on the configuration or use case.
An intelligent codebase processing server that provides agentic RAG capabilities for code repositories, enabling semantic search and contextual understanding through self-evaluating retrieval loops.
Related MCP Servers
- -securityFlicense-qualityA server exposing intelligent tools for enhancing RAG applications with entity extraction, query refinement, and relevance checking capabilities.Last updated -23Python
- -securityAlicense-qualityA server that integrates Retrieval-Augmented Generation (RAG) with the Model Control Protocol (MCP) to provide web search capabilities and document analysis for AI assistants.Last updated -1PythonApache 2.0
- -securityFlicense-qualityA local server that provides powerful code analysis and search capabilities for software projects, helping AI assistants and development tools understand codebases for tasks like code generation and refactoring.Last updated -2Python
- -securityAlicense-qualityAn intelligent server that provides semantic code search, domain-driven analysis, and advanced code understanding for large codebases using LLMs and vector embeddings.Last updated -3PythonMIT License