Desktop · Windows

Your AI handler,
right where your data lives.

MyHandler Desktop runs entirely on your Windows PC. Screen text, meetings, memories — captured, encrypted, and searched without leaving your machine.

Free for Windows 10/11. Cross-channel features require a MyHandler.ai subscription.
MyHandler Desktop chat interface showing a Local/Cloud memory toggle and conversation history
Built on whisper.cpp · llama.cpp · sqlite-vec open-source, auditable, running on millions of machines.
Privacy

We didn't add privacy.
We started with it.

These aren't features we could turn off. They're constraints we built around.

Functions with zero network

Disconnect your computer. Everything still works. No silent update checks, no telemetry, no anonymous analytics. Nothing leaves.

Your screen, but never as pixels

Other tools record your screen pixel by pixel. We refuse. Text comes from accessibility APIs only — the same ones screen readers use. Every captured snippet, transcript, and embedding is encrypted with AES-256-GCM, with the key in Windows Credential Manager.

Cloud is opt-in

Cross-channel features require an explicit click. We will never default cloud sync on. Disconnect any time — your local memory stays yours.

How it works

Three local engines.
Zero pixels stored.

1

Reads what's on screen

Captures text from windows you allow, using Windows accessibility APIs. Never takes a screenshot. Never stores a pixel.

2

Transcribes your meetings

A local whisper.cpp engine turns voice into searchable text. Audio never leaves your computer.

3

Remembers, on-device

A local llama.cpp engine and SQLite vector index let you search everything you've seen, said, and decided — semantically, instantly, offline.

Why we're different

Not Recall. Not Rewind. Not Littlebird.

Approach MyHandler Desktop Microsoft Recall Rewind Littlebird
Captures screen as Text via accessibility APIs Pixel screenshots Pixel screenshots + OCR Pixel screenshots
Local-only by default Cloud-assisted
Encrypted at rest (AES-256-GCM) BitLocker-dependent Disk-level only Cloud-side
Local LLM inference
Local meeting transcription
Connects to cross-channel handler
Free tier for the desktop app Limited

Comparison based on publicly documented behavior at the time of writing. Trademarks belong to their respective owners.

Optional cloud companion

Same handler. Two surfaces.

Connect MyHandler Desktop to a MyHandler.ai subscription, and your local context joins the same handler that screens your email, drafts replies, and answers your phone. The Desktop app stays local. Only the queries you choose to share cross the line.

Learn about the cloud handler →
Desktop app
Free

Local capture, meeting transcription, on-device search. Forever, no account required.

Cloud sync
Subscription

Email triage, voice screening, cross-channel handler. Requires a MyHandler.ai plan.

Requirements

Built for modern Windows.

  • Operating system Windows 10 (1809+) or Windows 11
  • CPU x64, 2019 or newer (Zen 2 / 10th-gen Intel or newer)
  • Memory 8 GB minimum, 16 GB recommended
  • Disk 50 MB installer; 3–5 GB after first-run model download
EST. 2026 — MYHANDLER.AI

A private AI handler,
installed on your machine.

Join the waitlist
macOS and Linux aren't supported in v1. Want them? Tell us when you join.
Frequently asked

Questions, answered.

Is MyHandler Desktop really free?

Yes. Download it, use it locally, forever. Cross-channel features (email triage, voice screening, the cloud handler) require a MyHandler.ai subscription.

How is this different from Microsoft Recall or Rewind?

We never store pixels. Screen capture uses Windows UI Automation — text only, from windows you've allowed. No images of your screen exist anywhere.

Does it send my data anywhere?

Not unless you turn cloud sync on. The app works fully offline by default. When cloud sync is enabled, only the specific queries you make leave the device.

What about macOS or Linux?

Windows-only at launch. Mac and Linux are on the roadmap but not committed. Vote with the waitlist.

Where do the AI models run?

Locally. We bundle whisper.cpp for transcription and llama.cpp for inference. Both run on your CPU/GPU. Nothing routes through us.