qmd
mini cli search engine for your docs, knowledge bases, meeting notes, whatever. Tracking current sota approaches while being all local
QMD is an MCP server that provides hybrid semantic and keyword search over local markdown documents, notes, and knowledge bases. It combines BM25 full-text search, vector embeddings, and LLM re-ranking to help AI agents find relevant information from indexed collections. The server runs entirely on-device using GGUF models via node-llama-cpp, requiring no external API calls. Developers can connect it to Claude Desktop, Claude Code, or other MCP clients to give agents persistent memory and retrieval capabilities across their personal documentation.
Key Features
Use Cases
- 01Search personal notes, meeting transcripts, and documentation from within Claude Desktop or Claude Code
- 02Build AI agents that retrieve relevant context from indexed knowledge bases before answering questions
- 03Enable LLMs to find code examples and API documentation from local repositories
- 04Create retrieval-augmented workflows where agents pull supporting documents based on user queries
- 05Index technical documentation and let agents answer questions with exact source citations
- 06Search across multiple collections (work docs, personal notes, project wikis) in a single query
Related MCP Servers
View moremarkitdown
Python tool for converting files and office documents to Markdown.
firecrawl
The context API to search, scrape, and interact with the web at scale. 🔥
prompts.chat
f.k.a. Awesome ChatGPT Prompts. Share, discover, and collect prompts from the community. Free and open source — self-host for your organization with complete privacy.
langflow
Langflow is a powerful tool for building and deploying AI-powered agents and workflows.
qmd — FAQ
What is QMD?+
QMD is an on-device search engine that indexes markdown files, notes, and documentation into a hybrid search system combining keyword matching, semantic vectors, and LLM re-ranking. It exposes an MCP server so AI agents can query and retrieve documents from your local collections.
How do I install QMD for Claude Desktop?+
Install globally with npm install -g @tobilu/qmd, then add the MCP server configuration to ~/Library/Application Support/Claude/claude_desktop_config.json with command 'qmd' and args ['mcp']. After installation, index your documents with qmd collection add and qmd embed.
Which AI clients work with QMD?+
QMD works with any MCP-compatible client including Claude Desktop and Claude Code. For Claude Code, you can install it as a plugin via the marketplace or configure it manually as an MCP server.
Do I need API keys or external services?+
No, QMD runs entirely locally using GGUF models downloaded from HuggingFace. The embedding, re-ranking, and query expansion models (totaling ~2GB) are cached locally and require no API keys or internet access after download.
What are the system requirements?+
QMD requires Node.js 22+ or Bun 1.0+. On macOS, you need Homebrew SQLite for extension support (brew install sqlite). Models run on CPU by default; GPU acceleration is available via Metal, CUDA, or Vulkan if detected.
Is QMD free to use?+
Yes, QMD is open-source under the MIT license. All models are freely available from HuggingFace and the tool has no subscription or usage fees.
How do I install qmd?+
Open the source repository on GitHub and follow its README. qmd is a mcp server — MCP Agents Market links you directly to the official repo.
Is qmd free?+
qmd is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.