Meeting Assistant.

For Mac · macOS 14+ · Notarized by Apple

The meeting is in their language. You read it in yours.

Meeting Assistant sits beside your Zoom, Meet, or Teams call, transcribes both sides on your Mac, translates the other side live — and after every turn of theirs, drafts an answer you can read aloud in their language.

Download for macOS Free 14-day trial · full features · no account
Listening · English → Русский
Them · 00:12 Не могли бы вы рассказать о сроках пилотного проекта?
You · 00:19 Да, конечно — мы уже закончили интеграцию. Yes, of course — we have already finished the integration.
Suggested reply · read this aloud
Мы можем начать пилот через две недели.

Privacy is the architecture, not a policy

Your audio and your documents never leave this Mac. Speech is transcribed on the device. Only the text being translated is sent — to the AI provider you configure, with your own key.

No account, no cloud recording, no server in the middle. Even the license check runs offline. The one network call the app makes on your behalf is the translation request to Anthropic, OpenAI, or Google — under your key, your billing, your control.

How a call goes

01 · Consent

Ask first, then listen

Before anything is captured, the app hands you a consent script in the other side's language. They agree — the session starts. They decline — nothing is recorded. The answer is kept for your records.

02 · Read along

Both sides, live, translated

Their voice comes from system audio, yours from the microphone. Everything is transcribed on-device — Apple's recognizer or local Whisper — and the other side's words settle into your language as they speak.

03 · Answer well

A reply, ready to say

After each thing they say — a question or not — a suggested answer appears in their language, with a gloss in yours, grounded in the documents you attached. Earlier suggestions stay in a scrollable history, and you choose what fires automatically: replies to questions, to everything, or translations only.

Reference materials

Your documents answer with you

Before the call, drop in the price list, the spec, the contract draft — folders work too. When a question lands, the suggested reply is grounded in what you attached: your numbers, your terms, quoted correctly while you're under pressure.

It works across languages — a question asked in English finds the answer inside a document written in Russian. And it's all indexed on your Mac: the files never move, and only the few sentences relevant to the question ever reach the model.

PDF, Word, Markdown, RTF, and plain text.

Materials · this meeting
pricing-2026.pdf integration-spec.docx contract-draft.md
Them · 14:32 And what would the license cost for twenty seats? А сколько будет стоить лицензия на двадцать мест?
Suggested reply · read this aloud
For twenty seats it comes to €14,400 a year — that already includes the volume discount from our 2026 pricing.
На двадцать мест выходит 14 400 € в год — уже с учётом скидки за объём.
grounded in pricing-2026.pdf · “Volume tiers”

Built for real working calls

Eight languages

English, Russian, Ukrainian, French, German, Spanish, Chinese, and Portuguese — for the meeting and for the interface itself.

On-device transcription

Apple's built-in recognizer starts instantly; local Whisper trades a one-time download for higher accuracy on accents and cross-talk.

Consent, built in

A ready consent script in the other side's language, asked before a single word is captured — and their answer kept for your records.

Bring your own AI

Works with your Anthropic, OpenAI, or Google key. Pick any model for translation and another for replies; a meeting costs pennies on your own billing.

Nothing to install around it

No virtual audio drivers, no browser extensions, no bots joining your call. One signed, notarized Mac app.

A license, not a login

One key, pasted once, verified on your Mac. Works on a plane. Works in ten years.

Pay once. Own it.

Trial
$0 14 days
  • Every feature, no account
  • Uses your own AI key
  • No card, no strings
Download the trial
Personal license
$59 one-time
  • Yours forever
  • Free updates for a year
  • All your Macs, one person
  • Team packs available
Get a license by email
Managed plan
$12 / month
  • No API key to manage
  • Usage allowance included
  • Same privacy: your Mac talks to the AI provider directly
Coming soon

Questions people ask

Do I need an API key?

For now, yes — a key from Anthropic, OpenAI, or Google, pasted once into Settings. It takes five minutes to get and a typical hour-long meeting costs a few cents on your own billing. The managed plan will remove this step entirely.

Does my audio go anywhere?

No. Speech recognition runs on your Mac — Apple's on-device engine or local Whisper. Only the text of what was said is sent to the AI provider you chose, for translation and reply drafting. Your documents are indexed locally; only the short passages relevant to a question are ever included in a request.

Is it okay to use this in a meeting?

The app is built consent-first: it asks the other side in their own language before anything is captured, and records their answer. Laws on recording calls vary by place — the consent flow is designed so you're always on the right side of the strictest ones, but the responsibility for using it lawfully is yours.

What do I need to run it?

A Mac on macOS 14 or newer. Apple Silicon and Intel are both supported. The app asks for microphone and screen-recording permission — the second one is how macOS lets an app hear the other side of a call.

Do I need a headset?

Strongly recommended. With a headset, the microphone hears only you, and the transcript stays perfectly clean. On open speakers, the other side's voice also reaches your microphone — the app filters the obvious duplicates, but a headset removes the problem entirely, same as in any call app.

Which meeting apps does it work with?

Any of them. It listens to what your Mac plays and what your microphone hears, so Zoom, Google Meet, Teams, Slack huddles, and plain phone calls through your Mac all work — no plugins, no bot participants.