Cuena — Undetectable AI for meetings and your screen

Listen

Perfect notes you didn't take.

Cuena writes the meeting down as it happens, who said what, the decisions, the follow-ups, then hands you a clean summary the moment the call ends. It captures system audio directly, so there's no bot to invite.

Live transcript with speaker labels
Instant summaries and action items
Every session stored locally on your device

Ask & Magic

The answer, before they finish the question.

Two ways to get help. Ask when you have a question, type it and Cuena answers using your screen and the conversation. Magic when you do not, it reads the moment and proactively hands you the next move.

Sees your screen plus the last minute of audio
Streams into the floating header, never mid-screen
Magic needs zero typing, one keystroke and done

Cuena answering a question about a launch plan and spreadsheet using screen and recent audio context. — Ask anything about your screen. Cuena reads the visible work and recent audio together.

What Cuena does

Four things, always one keystroke away.

It hears the room, answers about your screen, tells you what to say next, and remembers all of it, without ever getting in the way.

Listen

Real-time transcription

Live transcript with speaker labels and instant summaries in your language, so you can be present instead of scribbling notes.

Ask

Answers about your screen

Ask anything. Cuena sees your screen and the last minute of audio, then answers the question in front of you. No copy-paste, no tab-switching.

Magic

Knows what to say next

Zero input. Magic reads the whole conversation and your screen and proactively suggests the next line, answer or follow-up, before you have to ask.

Remember

Searchable local memory

Every session is saved to a private, searchable history on your device. Any name, number or decision is one search away, days later.

Cuena searchable local history showing meeting summaries, transcript search, and a found decision. — Search a name, number, decision or follow-up across every saved session.

Remember

Everything said becomes searchable context.

Cuena turns each session into private, reusable knowledge: transcripts, summaries, decisions, speaker labels and action items, all stored locally so you can recover details days later.

Find names, numbers, risks and owners across sessions
Review summaries with action items after the call ends
Export Markdown notes when you need to share or archive

SearchLocal DBExport

Modes

Create any mode for the work in front of you.

Build as many custom modes as you need. Start from a preset, then tune the prompt, tone, triggers and context.

Create as many modes as you need, or start from a preset and tune it.

Custom mode

Tell Cuena how to think for this task.

Create a mode for board prep, support triage, tutoring, coding, founder pitches, or anything else.

Roleboard prep coachTriggerinvestor Q&AContextscreen + audio

+ New modeDuplicateTune prompt

Preset starters

Start fast, then edit every instruction.

Sales calls

Objections, next steps, follow-up lines.

Interviews

Sharper answers and better follow-ups.

Screen solver

Work through code, formulas, and docs.

Privacy-focused

Only you can see it. Only you can read it.

Cuena is a floating desktop layer, not a bot on the call, not a plugin in the meeting. It stays out of every recording and screen share, and your content stays under your control.

Cuena private AI assistant visible only to the user on top of a desktop screen. — The private layer sits with you, while the shared screen stays clean.

Invisible to screen share

Cuena's windows are excluded from screen capture and recordings at the OS level, a privacy shield for salary talks, medical calls and sensitive reviews.

No bots, no announcements

It captures system audio directly, so there is no "Cuena Notetaker" on the guest list and no "recording started" ping. Quiet, but your notes stay yours.

Follows your eyes

A frameless pill you drag anywhere and summon with one keystroke, so answers land exactly where you are looking. Never mid-screen, never in the way.

To be clear: the privacy shield protects your sensitive moments; it is not meant to deceive anyone. Cuena helps you work better, it is not a tool for cheating in interviews or exams.

Nothing leaves your device

With your own key or local models, audio and screen captures are processed on-device. In local mode it works on a plane.

Never trained on your content

Your transcripts and screens are never used to train any model, and never sold. Managed requests send only the minimum.

Your data, in a file you can open

History lives in a plain local database on your machine, not locked in someone else's cloud you can't inspect.

Your model, your bill

Bring your own AI. Or run it all locally.

Cuena is model-neutral. Use the managed credits on any plan, or upgrade to Pro to plug in your own provider key and run local Whisper and Ollama fully on-device. No lock-in, no middleman on your data.

Managed credits, nothing to set up
BYOK on Pro, your account, your billing
Local Whisper and Ollama, included with Pro

Cuena runtime mode panel showing managed credits, bring your own key, and fully local Whisper and Ollama options. — Switch between managed, BYOK and local modes before you send the request.

OpenAI

GPT models

Anthropic

Claude

Google

Gemini

Deepgram

Speech-to-text

Whisper

On-device STT

Local

Ollama

On-device LLM

Local

Pricing

Start free. Upgrade when it earns its keep.

A free tier with 50 monthly credits, a Pro plan for your own keys and a bigger allowance, and one-time packs when you need more. Billed in USD, cancel anytime.

Free

$0 / forever

50 managed AI credits every month
Ask, Magic and live transcription with summaries
Searchable local history on your device
No payment details required

Download for free

Questions, answered.

What's the difference between Ask and Magic?

Ask is for when you have a question, type it and Cuena answers using your screen and the recent conversation. Magic is zero-input, it reads the whole conversation and your screen and proactively suggests what to say or do next, no typing required.

Is it really invisible in screen sharing?

Yes. Cuena's windows are excluded from screen capture and recordings at the OS level, so they don't appear in a shared screen or a meeting recording. It's a privacy shield for your sensitive moments, not a tool to deceive the other side.

Does it join the meeting or show up as a participant?

No. Cuena never joins your call. It captures system audio directly from your desktop, so there is no bot on the guest list and no "recording started" announcement.

Is my data private?

With your own provider key or local models, your audio and screen content stay on your device. Managed mode sends only the minimum needed per request, and your content is never used to train models or sold. Your history lives in a local database file you can open yourself.

Which AI models can I use?

OpenAI, Anthropic Claude, Google Gemini and Deepgram through managed credits, and on Pro you can bring your own key or run fully local Whisper and Ollama.

Do I have to pay? Which platforms are supported?

The free plan includes 50 managed AI credits every month. Pro is $15/mo and adds your own provider keys plus local Whisper and Ollama, with a bigger monthly credit allowance. macOS today; a Windows version is coming soon.

Your private co-pilot, ready on desktop.

Free to start: your account gets 50 managed AI credits every month. Download the desktop app and you are up and running in a couple of minutes.

Download for MacmacOS 11+ · Apple Silicon

Coming soonDownload for WindowsIn the works

On an Intel Mac? Get the Intel build. First launch needs a one-time step — the install guide walks you through it.