add-qmd

Replace grep-based memory search with local QMD hybrid search.

189|60|Updated Mar 22, 2026
One-click install
npx skills add https://github.com/sbusso/claudeclaw --skill add-qmd
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: add-qmd
Source: https://github.com/sbusso/claudeclaw/tree/main/skills/add-qmd
Command: npx skills add https://github.com/sbusso/claudeclaw --skill add-qmd

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Add QMD as an advanced memory search backend to ClaudeClaw, enabling local, hybrid search and on-device ranking without external APIs.

Core Features & Use Cases

  • Replaces grep-based memory_search with QMD's hybrid BM25 + vector search + LLM re-ranking for memory recall.
  • Keeps everything local (no API keys), indexes per-group memory, conversations, and CLAUDE.md.
  • Use case: enhance long-memory recall across large groups and conversations for better agent responses.

Quick Start

Install QMD MCP server, index all group memory, and switch ClaudeClaw to use QMD for memory search.

Frequently Asked Questions about add-qmd

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I improve agent memory recall with local semantic search?

Local semantic search improves memory recall by replacing grep-based lookups with a QMD backend. This Skill indexes per-group memory directories, CLAUDE.md, and conversations to provide hybrid BM25 and vector search with LLM re-ranking directly on-device.

Can I use QMD for memory search without external API keys?

Yes, QMD memory search operates entirely locally without external API keys. It performs on-device ranking using hybrid BM25, vector search, and LLM re-ranking to keep your memory archiving process private and self-contained.

What is the best way to search across multiple group memory directories?

The best way to search across multiple group memory directories is indexing them with a local QMD MCP server. It scales memory recall by indexing groups/*/memory, conversations, and CLAUDE.md files for comprehensive hybrid search.

How do I install and index QMD as an MCP server for ClaudeClaw?

To install QMD as an MCP server for ClaudeClaw, configure the server locally and index all group memory directories. Once indexed, switch ClaudeClaw to use the QMD memory search backend to enhance long-memory recall.

Does QMD memory search provide a fallback if the server is unavailable?

Yes, QMD memory search provides a graceful fallback. If the QMD server is unavailable, the system automatically reverts to the standard grep-based memory search to ensure continuous operation.

Why does local memory archiving need hybrid BM25 and vector search?

Local memory archiving needs hybrid BM25 and vector search to improve recall accuracy across large groups and conversations. Combining keyword matching with semantic vectors and LLM re-ranking yields better agent responses than grep alone.