vemry
中文Sign in

Turn what you read into what you know.

Capture a thought, clip a page or a paper, read it beside a translation and review it as cards. The AI does the chores; the reading and thinking stay yours, in plain Markdown on your computer.

Download vemryOther platforms

Free, for macOS, Windows and Linux.

A chapter under study: the paper's own paragraphs embedded in the note, your retelling, and a check that points out, item by item, where it drifts.

Focus on the writing. The rest is a keystroke away.

Formatting, linking and finding a command never take you off the line you're writing. Type / at the start of a line and headings, tasks, tables, code and callouts are in one menu that narrows as you type. ⌘P opens the command palette, where each of the app's eighty-odd commands can be found by name.

Rich text copied from the web pastes in as clean Markdown. However you write it, what's saved is always standard Markdown.

/
Insert any kind of block
[[
Link a note, or one of its sections or blocks
@
Dates, due dates, priorities and references
⌘P
Command palette
⌘O
Open a note by title
⌘⇧F
Search every note

Type / to insert a question callout, then find side-by-side translation in the command palette.

Catch the thought before it's gone

Whatever app you're in, press ⌥⇧Space and the capture box drops in. Type a line, press Enter, and it's in your inbox without pulling you out of what you were doing. When captures pile up, File the inbox sorts each one into the note where it belongs.

Pair your phone by scanning a code, and what occurs to you on the go lands in the same inbox: within a second if your computer is on, or the next time you open it if not. Without a connection, the phone holds on to it.

Phone capture needs a vemry account (free); your captures travel through it to your computer.

One thought captured from another app with a shortcut, one from a phone; both land in the same inbox.md.

All on your machine. You decide what the AI sees.

vemry needs no sign-up, sends no telemetry and works with the network off. Connect a model that runs on your own computer (Settings has an Ollama preset) and even the AI stays on your machine.

Using a cloud model? Draw your privacy zones first. Mark a folder, a note or even a single passage private, and the AI can't read it, can't change it and can't tell anything is missing: asking for a private file gets the same error, byte for byte, as asking for one that doesn't exist.

The private folder is marked private, then the AI is asked to list every note: the folder isn't in the answer, and nothing hints that anything is missing.

Stop getting stuck on the language

Open a page in the built-in browser and press Translate: each paragraph's translation appears beside it as you read. For a PDF paper, press ⌘⇧T: the page on the left, its translation on the right, each paragraph level with its original. Click a translation and the passage it came from lights up, so checking it takes one glance.

Formulas stay as they are, and captions and references are never sent for translation. Notes can be read side by side too, interleaved or as translation only, and a translation can be saved as a note of its own. Translate into English, Chinese, Japanese, Korean, French, German, Spanish and more.

A PDF translated paragraph by paragraph into your language; click a translation and the passage it came from lights up. Try another target language too.

Clip anything worth keeping

Press ⌘⇧S in the built-in browser and the page becomes a note in clips/, with its text, its images, its link and the date. The images are saved into your own folder, so the note outlives the page. Want only one passage? Select it and press Selection → note, and it lands in the note you're writing, with its source.

Papers are easier still: paste an arXiv link and every section, equation, figure and table becomes a block you can cite on its own. On macOS, the built-in browser also blocks ads and trackers.

A passage clipped into the note being written, with its source; then the whole page saved to clips/ with ⌘⇧S.

Understand it chapter by chapter. See it again before it fades.

To read a paper, pick Study a paper and paste the link. The source comes in as blocks and is split into chapters; each chapter starts from what your vault already knows, then moves on to what's new, with the passages it draws on embedded alongside.

Finish a chapter and retell it in your own words: you'll see what you got right, what you left out and which symbol you mixed up, each point linked to the paper. Everyday reading works the same way: open an article in the Reader, ask as you go, and every answer names its source.

Pick Study a paper, paste the link and approve the plan; each note the AI wants to write waits at a diff until you allow it.

Every change is shown to you first

Whenever the AI wants to change a note, it stops at a diff first: allow it, deny it, or edit it and then apply. A whole turn's changes can be undone at once, and every session is kept in an audit log you can replay.

Reviewing a card that's due: recall first, then show the answer and grade how hard it was.

What didn't stick becomes a card

Cards come only from the gaps in your retelling. The Review view brings them back as they start to fade: press Good if you remembered and the next review comes later, or Again if you didn't. The cards and their schedule live in your notes, as plain Markdown.

Plain Markdown, seen any way you need

vemry keeps no database and no format of its own. Your notes are .md files that read the same in any editor, and if you ever switch tools, you take the folder with you.

The same notes are to-dos in Agenda, progress on the Board, a history on the Timeline, connections in the Graph and cards in Review. Views are queries that store nothing of their own: tick off a task in any of them and what changes is that line in the file.

The file itself on the left, views of it on the right; tick a task on the Board and the line in the file changes with it.

A few things you may want to check first

  • The source is public

    All of it is on GitHub for anyone to read, under FSL-1.1-ALv2; each version becomes Apache-2.0 two years after its release.

  • No telemetry

    Without signing in there's no account, and the app collects and reports nothing about how you use it.

  • Every step on the record

    What the AI read, what it wrote and how you answered are all in the audit log, and Session Replay steps through them.

  • Signed updates

    Updates are signed with minisign against a key built into the app, and downloading and installing both wait for your yes.

  • Sixteen built-in skills

    Tidy a folder, file the inbox, start and close the day, draft the weekly report, turn meeting notes into actions. To add your own, write one Markdown file.

  • Scripts instead of plugins

    Drop a .js file into .vemry/scripts/ to add commands, react to events or run on a schedule, inside a sandbox with no access to files, the network or processes.

  • Answers with sources

    Every answer names the block it's based on, and with web search on, each source is marked verified or not.

  • Three platforms, two languages

    On macOS, Windows and Linux, with the interface in English or Simplified Chinese, light or dark to match your system.

Free to use. Pro if you'd rather skip the API key.

Free $0
The editor, every view, the built-in browser and clipping. Add your own API key and you get the full AI as well.
Pro$5 / month, or $50 / year
No key to set up: 3,000 credits of vemry hosted AI a month, enough for about 30 close readings.
Founder$4 / month, or $40 / year
For the first 500 accounts: Pro at 20% off for life, with exactly the same credits and features.

A new account's first sign-in from the desktop app comes with 600 credits to try, valid for 14 days, no card needed. Full pricing and questions

Pick a folder and start reading

vemry opens an ordinary folder on your computer. If it already holds Markdown notes, they work as they are.

Download vemryOther platforms

Free, for macOS, Windows and Linux.

Curious what a lived-in vault looks like? The source repository has an example vault you can open in vemry.