A local search and indexing CLI that combines BM25, vector search, and reranking with native MCP support for AI agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install max-qmd
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install max-qmd using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is qmd?
qmd is a high-performance command-line utility designed to index and search local files using advanced retrieval techniques. It bridges the gap between raw file systems and AI-driven workflows, making it a vital component for developers utilizing Openclaw Skills. By leveraging BM25 keyword matching, vector embeddings, and reranking, it provides highly relevant search results from your local markdown files and codebases. The tool also features a native Model Context Protocol (MCP) mode, allowing AI agents to interact directly with your indexed knowledge base without manual file traversal.
qmd Use Cases
- Creating a private, local knowledge base from technical documentation for AI agent reference.
- Improving code search accuracy by combining keyword and semantic vector search.
- Providing precise context to AI coding assistants using line-specific document retrieval.
- Integrating local file indexing into automated workflows via the Openclaw Skills ecosystem.
How qmd Works
- Initialize a collection by pointing the tool to a local directory and defining file masks for inclusion.
- Update the index to process files, generating both BM25 statistics and vector embeddings via a local Ollama instance.
- Execute search queries using keyword, vector, or hybrid modes to find relevant fragments.
- Use the retrieval command to extract specific lines or sections from the indexed documents.
- Launch in MCP mode to expose the search capabilities to AI agents for automated reasoning.
qmd Setup
To get started with qmd for Openclaw Skills, install the package via npm:
npm install -g https://github.com/tobi/qmd
Ensure Ollama is running locally for vector capabilities, then configure your first collection:
# Add a directory to index
qmd collection add ./docs --name my-docs --mask "**/*.md"
# Build the index
qmd update
# Test the search
qmd query "how to configure api keys"
qmd Data Schema & Taxonomy
qmd maintains its indices and metadata locally to ensure privacy and speed.
| Component | Detail |
|---|---|
| Cache Directory | ~/.cache/qmd |
| Embedding Engine | Ollama (Default: http://localhost:11434) |
| Search Algorithms | BM25, Vector (Cosine Similarity), Hybrid Reranking |
| Metadata | Collection names, file paths, and glob patterns |
This structure allows Openclaw Skills to efficiently query large datasets without cloud latency.
qmd Advanced Features
- Support for the Model Context Protocol (MCP) for seamless AI agent integration.
- Hybrid search functionality that combines the precision of BM25 with the semantic understanding of vectors.
- Configurable reranking using local LLMs to improve result quality.
- Fine-grained control over document retrieval with line-based fetching (
qmd get). - Multi-collection management for isolating different project knowledge bases.
SKILL.md
Loading
Related Openclaw Skills
Humanizer
A sophisticated text editing skill that identifies and removes robotic patterns to inject personality into AI-generated content.
Content Machine
An autonomous 24/7 affiliate content production pipeline that orchestrates research, writing, quality control, and deployment for niche sites.
Larry: Autonomous TikTok Slideshow Agent
An AI-driven agent that automates the research, creation, and posting of high-engagement TikTok slideshows for affiliate marketing.
Humanizer: AI Text Refinement Skill
A specialized writing editor skill that identifies and strips away robotic AI patterns to restore a natural, human voice to your content.
qmd
A high-performance local search and indexing CLI that combines BM25 keyword matching with vector-based semantic search and MCP support.
Multi Search Engine
A comprehensive search integration providing API-less access to 17 global and domestic engines with advanced filtering capabilities.
Featured*
