This MCP server implements Retrieval-Augmented Generation (RAG) by storing embeddings in a Qdrant vector database and enabling semantic search over indexed documents. It supports ingestion from diverse sources and formats (PDF, DOCX, TXT, MD, HTML content from URLs), along with web scraping and bulk directory imports. Embeddings can be generated locally through Ollama or remotely via OpenAI, giving flexible deployment options for offline or cloud-based workflows. The server exposes a Claude Desktop integration, allowing users to add documentation, index sources, and perform semantic queries directly from Claude. The architecture includes a run.py entry point, mcp_server, rag_engine, and modular embedding providers, all designed to run on macOS, Linux, and Windows with a Conda-based environment. Prerequisites include Python 3.11+, Conda, Qdrant, Ollama (or OpenAI), and Claude Desktop. Configuration is done via environment variables (QDRANT_URL, EMBEDDING_PROVIDER, OLLAMA_URL, OPENAI_API_KEY) and optional Qdrant Cloud settings. The project exposes API functions such as add_documentation, add_directory, search_documentation, and list_sources. The focus is on fast retrieval, flexible chunking, and secure handling of API keys.