Every note-taking tool makes a fundamental decision: what format stores your content? Most choose a proprietary database. Glassy chose Markdown. This is not a stylistic preference — it is an architectural decision that affects every layer of the product, from search to AI integration to data portability. For a deeper analysis of the lock-in implications, see Markdown vs Notion: the lock-in question.
Why Markdown wins
- Human-readable — Open a .md file in any text editor. You can read it. No database queries, no API calls, no proprietary renderer.
- Machine-parseable — Every programming language has a Markdown parser. LLMs natively produce and consume Markdown. MCP tools return Markdown.
- Portable — Move your notes to Obsidian, VS Code, GitHub, GitLab, Notion (via import), Bear, iA Writer, or any other tool that supports Markdown. No conversion needed.
- Version-controllable — Put your notes in a Git repository. Diff them. Branch them. Merge them. Try that with a proprietary database.
- Future-proof — Markdown has been around since 2004. It will outlive every note-taking app currently on the market. Your notes will be readable in 2050.
How Glassy uses Markdown
Every piece of content in Glassy — notes, bookmarks, captures, voice transcripts, Obsidian sync — is stored as Markdown with YAML frontmatter. The frontmatter carries metadata:
---
title: "Attention Is All You Need"
source: "https://arxiv.org/abs/1706.03762"
captured: "2026-06-14T14:32:00Z"
type: "research"
tags: ["machine-learning", "transformers", "attention"]
---This format is identical to what Obsidian uses. It is parseable by any static site generator (Hugo, Jekyll, Eleventy). It is readable in any text editor. And it is exactly what LLMs produce when you ask them to write notes.
Markdown and MCP
When Glassy's MCP server
returns search results to Claude, Cursor, Windsurf, OpenCode, Hermes, Openclaw, Pi, or any MCP-compatible
AI agent, the content is Markdown. The agent does not need to parse a proprietary format or convert from
JSON — it gets Markdown, which is its native output format. The integration works without any conversion
step. When the agent writes a note back to Glassy via
create_note, it
sends Markdown. No conversion in either direction.
Markdown and Obsidian
Obsidian is the most popular local Markdown editor. Glassy syncs bidirectionally with Obsidian vaults because both use the same format — Markdown with YAML frontmatter and wikilink syntax. No conversion, no export-import cycle, no data loss. See our sync guide and migration guide.
Markdown and search
Glassy's hybrid search works on Markdown directly. BM25 tokenizes the Markdown text. Vector embeddings are generated from the Markdown content. No pre-processing step strips formatting — the search index sees the same content you see. This means code blocks, headings, and lists all contribute to search relevance.
The cost of not using Markdown
Notion uses a proprietary block-based format. Exporting to Markdown is possible but lossy — databases, views, relations, and rollups do not survive the conversion. Evernote uses a proprietary database format with an ENEX export that loses formatting and attachments. Roam Research uses a proprietary JSON format. See Glassy vs Notion and Glassy vs Evernote.
The cost is not just migration difficulty. It is dependency. When your content is in a proprietary format, you are dependent on the vendor for access. If they change pricing, deprecate features, or shut down, your data is trapped. Markdown eliminates this risk entirely.