Hey Product Hunt! I’m Srishesh, and I built dyer.md because I kept running into the same wall: I wanted a note-taking app that was beautiful, private, AND could talk to AI agents without me manually copy-pasting context every time. The breakthrough moment was MCP (Model Context Protocol). Once I realized AI agents could have structured access to a knowledge base, I rebuilt the entire architecture around flat Markdown files with a built-in MCP server. Now Claude, Cursor, or any MCP-compatible agent can search my vault, read specific sections, even update notes — all through scoped Vault Keys that I control. A few things I’m proud of: Every note is a file. Not a row in SQLite. Not a block in a proprietary format. A .md file you can open in vim. Git is the database. Every save = atomic commit. You get full version history without a “version history” feature. Encryption is real. AES-GCM, client-side. I can’t read your encrypted notes. Neither can MCP. Neural View maps your knowledge semantically using local embeddings — nothing leaves your machine (TBD) I’d love your feedback. What’s missing? What would make you switch from your current setup?

Dyer.md

Your notes are the AI’s memory. Own both.

Your notes are the AI’s memory. Own both.