# PyRag

> PyRag is an MCP (Model Context Protocol) server that provides AI coding assistants with access to current, comprehensive Python library documentation. It eliminates the frustration of outdated examples and wrong information that slows down development.

- **Type:** MCP server
- **Install:** `agentstack add mcp-nateislas-pyrag`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [nateislas](https://agentstack.voostack.com/s/nateislas)
- **Installs:** 0
- **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [nateislas](https://github.com/nateislas)
- **Source:** https://github.com/nateislas/PyRag

## Install

```sh
agentstack add mcp-nateislas-pyrag
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# PyRAG - Production-Ready AI Documentation Assistant

> **Transform your AI coding experience with comprehensive, real-time Python documentation**

## 🚀 **NOW LIVE & PRODUCTION READY** 🚀

**Server URL**: https://PyRAG-MCP.fastmcp.app/mcp  
**11 comprehensive libraries** • **Multi-dimensional search** • **Real-time streaming**

**The Problem**: LLMs have outdated knowledge. When you're coding with AI assistants like Cursor, they often give you wrong, outdated, or incomplete information about Python libraries. This causes:
- ❌ **Frustrating errors** and failed code generation
- ❌ **Wasted time** fixing outdated examples
- ❌ **Increased costs** from repeated queries trying to get it right
- ❌ **Development slowdown** when your AI assistant hits a wall

**The Solution**: PyRAG is a production MCP (Model Context Protocol) server that gives your AI assistant access to **current, comprehensive Python documentation** with advanced multi-dimensional search capabilities. No more outdated information, no more getting stuck.

## 🚀 **What This Means for You**

### **Advanced Search Capabilities**
- **Multi-Dimensional Search**: Parallel searches across architecture, implementation, deployment, monitoring, security, and testing dimensions
- **AI-Optimized Responses**: 10,000+ character comprehensive responses with complete topic coverage
- **Real-Time Streaming**: Live progress updates for complex queries
- **Intelligent Query Analysis**: Automatically detects simple vs comprehensive query intent

### **Production-Ready Features**
- **11 Major Libraries**: FastAPI, Django, LangChain, Grafana, Pydantic, LlamaIndex, Streamlit, CrewAI, LangSmith, Ragas, Flask
- **Comprehensive Coverage**: API references, tutorials, examples, and best practices
- **Live Server**: Deployed and ready at https://PyRAG-MCP.fastmcp.app/mcp
- **Seamless Integration**: Works automatically with Cursor IDE and other MCP-compatible assistants

### **Real-World Examples**

**Before PyRAG**: Your AI assistant gives outdated information that causes errors:

```
You: "How do I handle authentication in FastAPI?"

AI Assistant: "Use the old security approach:
from fastapi.security import OAuth2PasswordBearer
# ... outdated code that doesn't work with current FastAPI versions
```

**With PyRAG**: Your AI assistant provides current, working information:

```
You: "How do I handle authentication in FastAPI?"

AI Assistant: "Here's the current best practice using FastAPI's built-in security features:

from fastapi import FastAPI, Depends, HTTPException, status
from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
from jose import JWTError, jwt

app = FastAPI()
security = HTTPBearer()

async def get_current_user(credentials: HTTPAuthorizationCredentials = Depends(security)):
    try:
        payload = jwt.decode(credentials.credentials, SECRET_KEY, algorithms=[ALGORITHM])
        return payload
    except JWTError:
        raise HTTPException(status_code=401, detail="Invalid token")

@app.get("/protected")
async def protected_route(current_user = Depends(get_current_user)):
    return {"message": "Access granted", "user": current_user}
```

## 📚 **11 Production Libraries**

Your AI assistant now has comprehensive documentation for **11 major Python libraries**:

### **🏆 Production Library Coverage**

| Library | Documents | Focus Area |
|---------|-----------|------------|
| **FastAPI** | 14,135 | Modern web APIs & async frameworks |
| **Grafana** | 9,358 | Data visualization & monitoring |
| **LangChain** | 9,044 | LLM application development |
| **Django** | 7,785 | Full-stack web development |
| **Pydantic** | 5,664 | Data validation & settings management |
| **LlamaIndex** | 3,537 | RAG & document retrieval systems |
| **Streamlit** | 1,807 | Rapid AI application prototyping |
| **CrewAI** | 1,395 | Multi-agent AI systems |
| **LangSmith** | 1,290 | LLM observability & debugging |
| **Ragas** | 492 | RAG evaluation & metrics |
| **Flask** | 55 | Lightweight web applications |

### **📊 Documentation Coverage**

- **Content Types**: API references, tutorials, examples, guides, changelogs
- **Real-Time Updates**: Continuous ingestion ensures current information
- **Multi-Dimensional Coverage**: Architecture, implementation, deployment, monitoring, security, testing

## 🎯 **How It Works**

1. **You ask your AI assistant** about Python libraries (just like normal)
2. **Your AI assistant connects** to the PyRAG MCP server via HTTPS
3. **PyRAG searches** comprehensive, up-to-date documentation
4. **You get better answers** with current information and examples

**Simple setup**: Just configure your MCP client to connect to the PyRAG server, then start asking questions!

> **Note**: Claude requires HTTPS connections for security. See [MCP HTTPS Setup](docs/mcp_https_setup.md) for configuration details.

## 🔍 **What You Can Ask**

Just ask your AI assistant normally about Python libraries:

### **API Questions**
- "How do I use pandas.read_csv() with custom delimiters?"
- "What are all the parameters for requests.Session()?"
- "How do I create a FastAPI endpoint with query parameters?"

### **Code Examples**
- "Show me examples of async/await in aiohttp"
- "How do I implement caching in FastAPI?"
- "Give me examples of pandas data manipulation"

### **Troubleshooting**
- "Why am I getting a ModuleNotFoundError with pandas?"
- "How do I handle memory issues with large datasets?"
- "What's the best way to structure a FastAPI project?"

### **Best Practices**
- "What are the recommended patterns for error handling in async code?"
- "How do I optimize performance in data processing?"
- "What are common pitfalls when using LangChain?"

## 🛠️ **Getting Started**

### **For Cursor IDE Users**

**Step 1: Configure MCP in Cursor**
1. Open Cursor IDE
2. Go to **Settings** → **Extensions** → **MCP**
3. Add PyRAG server configuration:
```json
{
  "mcp.servers": {
    "pyrag": {
      "command": "curl",
      "args": ["-X", "POST", "https://your-pyrag-server.com/mcp"],
      "env": {}
    }
  }
}
```

**Step 2: Start Using It**
- Open a Python file in Cursor
- Ask your AI assistant: "How do I use pandas.read_csv()?"
- Your assistant will automatically use PyRAG for current documentation!

### **For Other MCP-Compatible AI Assistants**
Configure your MCP client to connect to: `https://PyRAG-MCP.fastmcp.app/mcp`

## 🏗️ **Production Architecture**

PyRAG features a sophisticated multi-dimensional search system:
- **Intelligent Query Analysis**: LLM-powered intent detection and query expansion
- **Parallel Search Execution**: Simultaneous searches across 4-7 knowledge dimensions  
- **Topic Coverage Engine**: Ensures comprehensive responses with gap detection
- **Real-Time Streaming**: FastMCP streaming with live progress updates
- **ChromaDB Cloud**: Production vector storage with comprehensive documentation
- **Crawl4AI Integration**: Unlimited local web scraping for current documentation

## 🤝 **Contributing**

We welcome contributions! See our [Contributing Guide](CONTRIBUTING.md) for details.

### **For Developers**
- **Data Ingestion Pipeline**: [Pipeline Overview](docs/data-ingestion-pipeline-overview.md)
- **RAG Architecture**: [RAG Pipeline Overview](docs/rag-pipeline-overview.md)
- **Multi-Dimensional Search**: [Search Engine Implementation](src/pyrag/search/)

## 📄 **License**

This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.

## 🙏 **Acknowledgments**

- **Crawl4AI** for unlimited local web scraping capabilities
- **ChromaDB Cloud** for production-grade vector storage
- **FastMCP** for streaming MCP server implementation
- **MCP Community** for the Model Context Protocol specification

## 📞 **Support**

- **Issues**: [GitHub Issues](https://github.com/nateislas/PyRag/issues)
- **Discussions**: [GitHub Discussions](https://github.com/nateislas/PyRag/discussions)
- **Documentation**: [Technical Documentation](docs/DOCUMENTATION.md)

---

**Ready to supercharge your AI coding experience?** 🚀

Connect to https://PyRAG-MCP.fastmcp.app/mcp and experience multi-dimensional search with comprehensive, production-ready responses that actually help you build better software faster!

## Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [nateislas](https://github.com/nateislas)
- **Source:** [nateislas/PyRag](https://github.com/nateislas/PyRag)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** yes
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/mcp-nateislas-pyrag
- Seller: https://agentstack.voostack.com/s/nateislas
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.