mdstill

LLM-ready document preprocessing.Any format → clean markdown → efficient prompt

by Myrtie McClure

Visit

Apr 28, 2026 AI & Machine Learning

document preprocessing llm workflow markdown conversion

Gallery

About

mdstill is a document-ingestion tool purpose-built for LLM and RAG workflows. Where generic converters dump messy text, mdstill outputs clean, semantic markdown that preserves tables, headings, and document structure — the things LLMs actually need to understand context.What you can do with it:Prepare documents for RAG pipelines (chunk-ready, semantic boundaries preserved)Feed PDFs, Word files, or spreadsheets into ChatGPT, Claude, or Gemini without losing tablesBuild knowledge bases in Obsidian, Notion, or Logseq from existing document archivesExtract structured context for AI agents and embeddingsHow it's different: Deep-conversion mode runs layout-aware parsing (tables, OCR, multi-column PDFs) — not just text dumping. Markdown output is ~40% more token-efficient than raw text, so your LLM costs drop. REST API available for pipeline automation.Free tier, no signup required for basic use. Competes with markitdown, Unstructured.io, and LlamaParse — but with a zero-friction web UI.