“Draft a reply to Maria about tomorrow’s ship.”
Meet Friday.
A voice assistant with hands on your Mac.
Draft email, book meetings, open projects, read your screen — just by saying what you need.
Watch Friday do its thing.
A real session — voice in, action out. No cuts.
Talk to it like a person.
Press one key and speak. Friday handles the way humans actually talk — false starts, self-corrections, half-formed thoughts, topic switches. No wake word to train. No rigid commands to memorise. Voice to transcript in under 300 ms.
One assistant. Every model.
Claude Sonnet and Opus for reasoning. Haiku for speed. Gemini Flash for vision and real-time decisions. Friday routes each task to the right model — and uses what’s on your screen as context. You never have to describe what you’re looking at, or remember which AI does what.
It actually does the work.
Friday drafts email in Mail, books events in Calendar, sends messages, silences notifications, opens apps, moves files, and reads what’s on your screen. One spoken sentence replaces minutes of clicking — you stay focused on what matters.
Built on native macOS APIs — EventKit, Contacts, AppleScript, ScreenCaptureKit — the same ones Apple’s own apps use. No browser plugin. No screen-scraping. No web wrapper.
"Draft a reply to Maria about tomorrow's ship, then book a 10am block to finish the fix."
Build failed on a Swift 6 concurrency warning. Nothing blocking. I'll push the fix tomorrow morning and we'll be green by 10.
Fits your workflow.
Friday plugs into the apps you already use — macOS native, productivity, dev tools. And any MCP server you connect: an open standard, no proprietary lock-in. No migration. Nothing new to learn.
Calendar
“Book 30 min with Jordan tomorrow at 10.”
Messages
“Tell Anna I’ll be 10 minutes late.”
Contacts
“What’s Marco Conti’s email?”
Notes & Reminders
“Remind me to review the PR at 3pm.”
Finder & Spotlight
“Open the Q2 roadmap in Downloads.”
Safari
“Search Apple Docs for ScreenCaptureKit.”
Slack
“DM the team: build is green, shipping at 10.”
Notion
“Create a meeting note in the Product database.”
Linear
“File a bug: AudioEngine crashes on Swift 6.”
Figma
“What’s the spacing token for primary buttons?”
GitHub
“Open last week’s PR review on AudioEngine.”
Xcode
“Run the last failing test and show me the diff.”
Terminal & shell
“Run the build script and send me the output.”
Raycast
“Trigger my clipboard-history extension.”
Anthropic Claude
Opus for depth, Sonnet for balance, Haiku for speed.
Google Gemini
Flash for vision and real-time decisions.
Soniox
Low-latency speech-to-text with Whisper fallback.
ElevenLabs
Realtime Flash 2.5 voice for Friday’s replies.
Any MCP server
Open standard. Connect your own tools, internal APIs, databases.
Your keys. Your data. Your AI.
Bring your own keys
You hold the Anthropic, Google, ElevenLabs and Soniox keys. You pay providers directly. We never touch your tokens, and you keep whatever billing relationships you already have.
Local-first by design
Context, memory, and configs live encrypted in your macOS Keychain. Your voice streams from your Mac to the providers you chose — never through our servers.
Swap the brain
Claude Opus for depth. Haiku for speed. Gemini Flash for real-time. You pick per task — Friday switches on the fly. New model releases? You change one setting.
No lock-in, ever
Cancel anytime, export anytime. Because your keys and data never leave your Mac, walking away costs you nothing. BYOK means zero switching cost, built in.
From nothing to working, in 60 seconds.
-
01
~20 s
Install
Drag Friday into your Applications folder. Notarized macOS app, no sandbox escapes, no kernel extensions.
-
02
~30 s
Paste your keys
Anthropic, Google, ElevenLabs, Soniox. Pasted once into the macOS Keychain. Skip any provider you don’t use — Friday routes around it.
-
03
immediate
Press & speak
Hold your push-to-talk key. Say what you need done. Friday handles the rest on your Mac.
Be first when Friday ships.
Day-one access when we launch. Priority invites throughout the rollout. One email when we ship — that's it.
Straight answers.
How quickly can I actually start using it?
About 60 seconds from first launch: install the app, paste the API keys you already have from Anthropic and Google into the Keychain dialog, assign a push-to-talk key, start talking. No account to create, no cloud sync to wait for, no onboarding wizard.
What does Friday actually do on my Mac?
It drafts email in Mail, books events in Calendar, sends messages in Messages and Slack, opens apps and files, moves things in Finder, reads what’s on your screen via ScreenCaptureKit, and executes multi-step tasks through MCP servers. One spoken sentence can replace a chain of clicks — it acts, it doesn’t just describe what you could do.
How is my data handled? Who sees what?
Your voice goes from your Mac directly to the AI providers whose keys you pasted (Anthropic, Google, Soniox, ElevenLabs) — never through our servers. Context, memory and configs stay encrypted in your macOS Keychain. We don’t run a backend that holds your data. We can’t see your conversations even if we wanted to.
Why Mac only?
macOS exposes the OS-level APIs (EventKit, Contacts, Accessibility, ScreenCaptureKit, AppleScript) that let a voice agent actually do things natively. Windows and Linux don’t have an equivalent shared surface area. We could wrap a browser on other platforms, but that’s a demo, not a product.
Does Friday speak languages other than English?
At launch, English only. Soniox and Whisper both support dozens of languages under the hood, so multi-language support is a configuration switch for v2 once we hear real demand. Tell us in the waitlist email which language you need and we’ll prioritise accordingly.
When does Friday launch?
Q2 2026. Waitlist members get day-one access and priority invites throughout the rollout. We’ll send one email when we ship — that’s the only one you’ll get from us until then.