Download

Maple for your desktop

A lightweight, local-first AI memory layer that lives in your menu bar. Captures meetings, indexes your screen, and gives you instant recall — without leaving your flow.

v0.9 Beta · Early Access
RECOMMENDED
macOS

macOS

Apple Silicon & Intel
macOS 13 Ventura or later
Universal binary · ~85 MB
Windows

Windows

Windows 10 & 11
x64 architecture
MSIX installer · ~90 MB
iOS Android

Mobile

iOS & Android
Companion app for on-the-go
recall and meeting prep
On Roadmap
What's included

Everything in one install

No plugins, no browser extensions, no manual setup. Maple ships as a single native app that does it all.

Meeting transcription

Auto-joins Google Meet, Zoom, Teams, and Slack calls. Transcribes locally with Whisper and generates summaries, action items, and follow-ups.

Screen context capture

Silently indexes what's on your screen using on-device OCR. Builds a searchable memory of everything you've seen — documents, tabs, Slack threads.

Instant recall

Press ⌥ Space to summon Maple from any app. Ask natural-language questions about your meetings, screen history, or connected tools.

Integration hub

Connects with Gmail, Google Calendar, Slack, Notion, and more. Pulls in context automatically — no copy-paste, no tab switching.

Proactive routines

AI-powered automations that surface relevant context before meetings, draft follow-up emails, and flag tasks you might have missed.

Private by design

AES-256 encrypted on-device storage. Indexing happens locally — only the snippets needed for a query are sent encrypted, and never used for training.

Up and running in minutes

No configuration wizards, no API keys. Install, grant permissions, and Maple starts learning.

1

Install

Download the native app for your OS. Drag to Applications on macOS or run the MSIX installer on Windows.

2

Grant permissions

Allow screen recording and microphone access. Maple requests only what it needs — nothing more.

3

Connect your tools

One-click OAuth for Gmail, Calendar, Slack, and Notion. Maple begins pulling context automatically.

4

Start asking

Press ⌥ Space from any app. Within hours, Maple has enough context to feel like a second brain.

System requirements

Maple is built with Tauri — lightweight, fast, and native to your OS.

macOS

  • Processor Apple Silicon (M1+) or Intel x86_64
  • RAM 8 GB minimum · 16 GB recommended
  • OS macOS 13 Ventura or later
  • Storage ~200 MB app + local index
  • Audio Built-in mic or external for transcription

Windows

  • Processor x64 processor (Intel / AMD)
  • RAM 8 GB minimum · 16 GB recommended
  • OS Windows 10 (1809+) or Windows 11
  • Storage ~250 MB app + local index
  • Audio Any audio input device

Built different

Maple isn't another Electron app wrapping a web page. It's engineered for performance, privacy, and peace of mind.

Tauri + Rust core

Native system tray app built with Tauri. Sub-50 MB binary with near-zero idle CPU usage. No bundled Chromium.

SQLite + local vectors

Your memory index lives in a local SQLite database with embedded vector search. Sub-100ms retrieval at any scale.

AES-256 at rest

All local data is encrypted with AES-256. Your workspace key never leaves your machine — zero-knowledge by default.

Whisper transcription

On-device Whisper model for real-time meeting transcription. No audio ever uploaded — transcripts stay local.

On-device OCR

Screen content captured via native OCR APIs. Continuous indexing that doesn't slow down your machine.

Auto-update

Silent background updates that apply on next launch. Always on the latest version — no manual downloads after setup.

Common questions

Maple is currently in closed beta. We're onboarding waitlist members in weekly batches. Join the waitlist and we'll email you when your batch opens — typically within 1–2 weeks.

Yes — the Free plan includes meeting and screen context capture, 3 active projects, and 5 hours of transcription per month. Pro unlocks unlimited projects, advanced AI, and integrations for $16/month.

Indexing happens entirely on your device. When you ask a question, only the minimal context snippets needed are sent — encrypted — to generate a response. Your data is never used for model training.

The index is highly compressed — typically 1–3 GB for a month of active use. You can configure retention windows and exclude specific apps from indexing in Settings.

Maple is built on Tauri with a Rust backend — it's designed for near-zero idle CPU and low memory usage. Screen capture and OCR run on a throttled background thread that pauses automatically during heavy workloads.

On macOS, drag Maple to Trash. On Windows, uninstall from Settings → Apps. All local data including your encrypted index is removed — nothing persists after uninstall.

Early access

Ready to remember everything?

Join thousands of knowledge workers building a persistent memory layer for their work.

No credit card required · Free plan available · Weekly batch onboarding