Early access — 50 spots left

Speak your prompts.
They appear instantly across your apps.

Stop copying and pasting between tools. Just talk — and your text shows up exactly where you need it.

Speak → instant textWorks across every appNo more copy-paste

If you use AI daily, this will save you hours.

Download for Windows — FreeSee how it works

Windows 10+ • No account required • No setup headaches

Windows will show a security prompt on install — code signing is in progress. We ship fixes fast.

Version 3.1.0
71 updates shipped
12 versions shipped
Active development

Every prompt costs you more than you think.

1Write your prompt somewhere
2Copy it
3Switch to the right tab
4Paste. Lose your train of thought.

Marcos removes all of that. You just talk — and it's already there.

How it works

Three steps. No learning curve.

01

Download & install

One installer. Runs in under a minute. Marcos appears in your system tray, ready to go.

02

Press Ctrl+Shift+M

A small animated mascot pops up on your screen. It's listening. Start talking naturally.

03

Your words appear instantly

Whatever app has focus gets the text. ChatGPT, VS Code, Slack, your browser — anywhere.

What you get

A voice-controlled layer for your whole workflow.

<300ms

Real-time transcription

Your words appear as you speak them — not after a pause. Feels like actual typing.

Works in every app

No plugins, no integrations. Text lands wherever your cursor is — any app, any window.

Works offline too

No internet? No problem. Marcos can run a local model entirely on your machine. Your audio never leaves your device.

Stays out of your way

Auto-hides when you're not using it. Appears when you need it. Lives quietly in your tray.

Your data stays yours

Nothing stored on our end. What you say is yours. Choose offline mode for complete privacy.

No clipboard interference

Text is typed directly as keystrokes. Your clipboard is never touched.

Built for people who live inside AI tools.

Prompt engineers & developers

You're writing to ChatGPT, Cursor, or Claude dozens of times a day. Speaking is faster and your hands stay on the keyboard where they belong.

Writers & content creators

You think faster than you type. Marcos keeps up. Speak a paragraph, edit later — your hands are never the bottleneck again.

Power users & builders

Slack, email, Notion, Jira, browser tabs. If you're moving text between apps all day, Marcos eliminates the friction entirely.

First of its kind

Offline voice-to-text.
No cloud. No API. No internet.

Every other voice tool sends your audio to a server. Deepgram, Whisper API, Google — all cloud-dependent. Marcos is the first dictation tool that runs a full speech model entirely on your machine. Your words never leave your device. Not even for a millisecond.

Every other tool
  • Audio sent to external servers
  • Requires internet connection
  • API key needed
  • Someone else has your voice data
Marcos (offline mode)
  • Runs 100% on your machine
  • Works with no internet
  • No account, no API key
  • Zero audio leaves your device

Powered by Whisper running locally. <300ms latency. No subscription required.

Your crew

Six characters. Pick your vibe.

Marcos is the default, but you can switch to Gerald, Fluffy, Leon, Panda, or the Owl. All animated. All audio-reactive.

MarcosGeraldFluffyLeonPandaOwl

Recently shipped

Full changelog →
v3.1.0Auto-hide, offline download progress, settings persistenceLatest
v3.0.0User accounts, dashboard overhaul, managed API key optionRecent
v2.1.2Offline speech-to-text — no internet, no API key requiredRecent

Get in early. Shape what it becomes.

Free while we're in early access. No account needed to download.

Early Access — Founding Users
Free

Download now. No credit card. No catch.

  • Full access to the current version
  • Direct influence on the roadmap
  • Priority access to future updates
  • Free before we ever charge
Download for Windows

Windows 10+ • ~30MB • You'll need to log in inside the app

Premium
Full Platform Access
$80 / month

The complete Marcos experience — desktop, mobile, and agent-powered.

  • Everything in Founding Supporter
  • iOS & Android app (coming soon)
  • Agent skills — voice-triggered multi-step sequences
  • Cross-app automation (move between apps, fill forms, execute tasks)
  • Voice-triggered calls & real-time interactions
  • Priority support & early feature access
Get Premium Access

Reach out — we'll set you up within 24 hours.

Common questions.

Is Marcos really free?

Yes — for early access users. The first 50 people get full access to the current Windows version at no cost, locked in before we ever start charging.

Why does Windows show a security warning?

Windows SmartScreen flags installers that aren't code-signed yet. We're in the process of getting a code signing certificate. You can safely click “More info → Run anyway.” The app is safe — we're just early. We ship fixes fast.

Does it work without internet?

Yes. Marcos has a built-in offline mode powered by Whisper running locally on your machine. Your audio never leaves your device — not even for a millisecond.

Which apps does it work in?

Any app with a text input — ChatGPT, Claude, Cursor, VS Code, Slack, Notion, Jira, your browser, email clients, anything. Marcos types directly as keystrokes, so if you can type somewhere, Marcos can speak there.

Why is Premium priced at $80/month?

Because this is a genuinely new kind of product. We're not just adding a microphone button — we're building a voice-native layer that sits across your entire computer, phone, and eventually every interface you use. iOS, Android, agent skills, real-time cross-app actions — this is infrastructure, not a widget. The $80 tier exists for early adopters who want to be part of something that could fundamentally change how people interact with computers and AI. More features are coming fast, and this price reflects the vision, not just what's already here. Rheo Research's goal is to advance how humans interact with technology across every platform — and Premium users are backing that directly.

What is Rheo Research?

Rheo Research is the company building Marcos. Our focus is advancing human-computer interaction — making interfaces more fluid, more intelligent, and more natural across desktop, mobile, and beyond. Marcos is the first product that embodies that mission. Learn more at rheoresearch.org.

Can I get a refund?

If you're not happy, reach out within 7 days and we'll sort it. No questions asked. Contact us here.

The roadmap is shaped by real usage.

These aren't promises. They're directions I'm actively thinking about — and your feedback moves them up or down.

Screen snipping

Select a region of your screen and attach it directly to your prompt — no screenshots, no drag-and-drop.

Agent actions

Voice-triggered sequences — “summarize this page”, “open Cursor and start a new file” — without touching the keyboard.

Floating mascot (optional)

An always-visible companion on your desktop — reactive to your mic, dismissible, customizable. Purely optional.

A quick heads up

This is an early version. Windows may show a security warning on install because the app isn't code-signed yet. Totally fair if that's a concern — I'm working on it.

Built in the open — feedback directly shapes the product.

This is being shaped directly by early users. If you try it, I'd genuinely love your feedback — what works, what doesn't, and what one annoying step in your AI workflow you'd most want to eliminate.

Questions? Want to back the project?
Say hello.

Whether you want early access, have feedback, or want to discuss the Founding Supporter tier — this goes straight to the founder.