Read what AI writes.
Keep what you think.

AI handles the writing.
This handles the reading.

Preview and source, one key apart. Typora-compatible themes work as they are. Mermaid, Graphviz and math, all tuned. The whole app installs at ~19 MB, with no browser engine inside.

A shaky line? Highlight it. A doubt? In the margin. Wrong? Fix it on the spot.

Claude, Codex and OpenClaw all have a chat window. None of them is made for reading.

The best of the last
generation. Built in.

local-first · git sync · outliner · [[wikilinks]] · backlinks · wiki pages · daily notes · plugins

Roam and Obsidian worked this out years ago. note.md puts it back into files. One plugin brings your whole Roam graph over. An Obsidian vault just opens.

Agent-ready by design.
Use the AI you already have.

Built-in agent workflows use the agents, AI subscriptions, API accounts, or local runtimes you already have. note.md sells no tokens, adds no token markup, and charges no separate per-token fee.

Model usage still follows your chosen provider's plan, limits, and billing — or your own local compute.

Your folder is the workspace every AI tool shares. Cowork, Claude Code, Codex, DeepSeek, ChatGPT, OpenClaw, Hermes — same files, same house rules, all in git. See how.

Every file it writes follows Open Knowledge Format v0.2, strictly: what a document is, where it came from, who checked it — plain YAML at the top of the file. Any tool can read your vault. Not just this one.

Built-in agents, without a second token meter. Change AI tools whenever you want. What's yours stays yours.

Need something else?
Add it.

Write a plugin. Add a scheduled job. Hang your skills off it.

Put a ? in a sidecar note and Smart Lookup writes a fenced answer beneath it — async, without touching the source document. You decide whether to adopt it.

Agents notice.
You decide what becomes memory.

Your preferences, boundaries and decisions should not be guessed. An agent finds candidates in the work and conversations you choose to bring in; note.md gives them back to you one at a time. Only what you confirm becomes lasting context.

Search finds what you wrote. Memory governs what an agent may believe about you — and what it may act on. note.md keeps those jobs separate on purpose.

Why trustworthy personal memory needs your yes →

Read in the right view.

Navigate long files with the table of contents, browse structured notes as Timeline or Index views, explore links in Knowledge Browser, and page through *.typeset.md books.

Bring more into the vault.

Sync Apple Notes and Roam Research, archive Meetings, and keep Assistant Mail as local records. Availability depends on platform; current Apple Notes and other native packages may be macOS-only.

Think spatially.

Open JSON Canvas files, format JSON, keep imported sources read-only, and let untitled notes and canvases receive stable names when you save.

Agents propose. You approve.

Smart Lookup returns governed answers, Memory applies role and scope, Conversation Transcript Corrections keeps human approval in the loop, and plugin installs hot-reload without restarting the app.