Every OpenClaw session starts with the same problem: your entire MEMORY.md gets injected into the API call. Every single turn. That\u2019s 2,000+ tokens of context \u2014 whether the conversation needs it or not.<\/p>\n
QMD fixes this. Built by Tobi Lutke (Shopify\u2019s CEO), QMD replaces brute-force memory injection with local hybrid search. It combines BM25 keyword matching, vector embeddings, and LLM reranking to retrieve only the 3-5 snippets that actually matter for each query.<\/p>\n
The results from our QMD local search setup: 90% fewer memory tokens per session<\/strong>, 47ms average search latency<\/strong>, and zero data leaving our machine<\/strong>. Here\u2019s exactly how to install, configure, and verify it. (see Google’s SEO Starter Guide<\/a>)<\/p>\n Without QMD, OpenClaw\u2019s memory system works like a sledgehammer. It reads your full MEMORY.md file \u2014 plus any daily memory files \u2014 and dumps all of it into every conversation turn. If your memory files total 2,000 tokens, you\u2019re burning 2,000 tokens on context every single time you send a message.<\/p>\n With QMD, that same memory gets indexed locally. When you ask a question, QMD searches your memory files and returns only the relevant snippets \u2014 typically 3-5 matches totaling around 200 tokens.<\/p>\n MEMORY TOKEN REDUCTION<\/p>\n 90%<\/p>\n ~2,000 tokens \u2192 ~200 tokens per turn<\/p>\n<\/div>\n Here\u2019s what makes QMD different from cloud-based RAG solutions:<\/p>\n The hybrid approach matters. BM25 alone misses semantic connections. Vector search alone misses exact keyword matches. QMD runs both, then uses a lightweight LLM reranker to sort the combined results by relevance.<\/p>\n Before starting your QMD local search setup, you need a few things in place. The biggest requirement \u2014 and the one most people trip over \u2014 is WSL2.<\/p>\n ⚠️ Warning<\/p>\n QMD will NOT install on native Windows. The npm package depends on a native sqlite-vec binary that doesn\u2019t compile on Windows. You must use WSL2 (Windows Subsystem for Linux).<\/p>\n<\/div>\n What you need:<\/strong><\/p>\nWhat QMD Does (And Why You Need It)<\/h2>\n
\n
Prerequisites<\/h2>\n