Memoe vs Handy

Jude Mercer, SEO and Content, Memoe — 29 October 2026

Memoe is an agent you talk to, and it talks back. It lives in your Mac's notch instead of in a chat window, so you can talk to it from whatever app you're in. It runs on the AI you already pay for — your Claude subscription through Claude Code, or your ChatGPT plan — so there's no new model to buy. Underneath, it runs Hermes Agent, the open-source agent harness from Nous Research, and it keeps a memory of you in plain files on your Mac that you can read, edit and delete.

Hold one key from any app, say what you need, and Moe answers out loud and goes and does it. It connects to around 1,400 services, and it has its own browser, signed in as you. You can message Moe over Telegram, WhatsApp, Slack, Discord, email, Signal and LINE. Before Moe sends anything to another person it shows you exactly what will go out and waits for your click. About 2 to 5 seconds before it speaks.

What Handy is

Handy describes itself on its own homepage as “speak into any text field,” and “the free and open source app for speech to text.” Its own video walkthrough explains the mechanism: “Simply press a keyboard shortcut, speak, and release, and Handy will paste whatever you said into the text field you're typing into.” Push-to-talk is its default mode — hold the key, speak, release — with a toggle mode as an alternative, where pressing the shortcut once starts recording and pressing it again stops it and transcribes; its docs set the default shortcut as Option+Space, note the key can be remapped, and add that “on macOS, a lot of people like using the fn key.” Its four homepage feature cards state: “Free — Accessibility tooling belongs in everyone's hands, not behind a paywall”; “Open Source — Together we can build further. Extend Handy for yourself and contribute to something bigger”; “Private — Your voice stays on your computer. Get transcriptions without sending audio to the cloud”; and “Simple — One tool, one job. Transcribe what you say and put it into a text box.” Its privacy policy adds that speech recognition “runs on your device,” that its models — among them Parakeet V3, Whisper and Cohere Transcribe, per its docs — “run entirely on your device,” and that during ordinary transcription Handy “does not send your microphone audio or transcription content to the Handy maintainers or to Handy servers.” It offers optional post-processing through a provider that “may include OpenAI, Anthropic, OpenRouter, Groq, Cerebras, Z.AI, and AWS Bedrock,” which is “disabled unless you configure and invoke it” and then sends the transcription, the selected prompt and the model information “directly from your device to the provider you selected,” with Handy's own maintainers not receiving or proxying that content, by its own account. Its download page lists builds for macOS (Apple Silicon or Intel, as a .dmg, or through a community-maintained Homebrew cask, per its docs), Windows and Linux, at version 0.9.8. Everything in this paragraph and in the right-hand column below comes from handy.computer, handy.computer/docs, handy.computer/download and handy.computer/privacy — their own home page, docs, download page and privacy policy — read 29 October 2026. Its privacy policy's effective date is given on that page as September 18, 2026.

Side by side

Memoe's column is the live, Approved wording from docs.memoe.app today.

	Memoe	Handy
What it is	An agent you talk to from any app, that answers out loud and does the work.	“Speak into any text field,” by its own homepage headline — “the free and open source app for speech to text.”
Starting a turn	Hold one key from any app and say what you need.	Press a keyboard shortcut, speak, and release, by its own homepage video — push-to-talk by default, with a toggle mode where pressing the shortcut again stops it, and the key remappable.
What comes back	A spoken answer, and the task carried out — mail and calendar read, a meeting noted, a Mac app driven for you.	Transcribed text pasted into the field you were typing into, by its own feature list: “One tool, one job. Transcribe what you say and put it into a text box.”
Where your speech is processed	“Speech recognition runs on your Mac using local models, and audio is not stored after it is transcribed. On the free tier your voice does not leave your Mac.”	“Your voice stays on your computer. Get transcriptions without sending audio to the cloud,” by its own homepage; its privacy policy adds that its models “run entirely on your device.”
The thinking	Your own Claude or ChatGPT plan, an API key, or a model on your Mac, driving Hermes Agent. Not resold, not marked up.	None by default. Optional cloud post-processing through a provider that “may include OpenAI, Anthropic, OpenRouter, Groq, Cerebras, Z.AI, and AWS Bedrock” is “disabled unless you configure and invoke it,” and then runs “directly from your device to the provider you selected,” by its own privacy policy.
What it runs on	An Apple silicon Mac, macOS 15 or later.	macOS (Apple Silicon or Intel, as a .dmg or through Homebrew), Windows, and Linux, by its own download page, at version 0.9.8.
Price	Free on the Mac. Away, from $20 a month, is a cloud computer of your own that keeps going when the Mac is shut; Away+ is $50 a month.	“Free,” by its own homepage badge, which adds: “Accessibility tooling belongs in everyone's hands, not behind a paywall.” Kept running by named sponsors and donations, by its own homepage.

Trying Memoe

Memoe is a free Mac app, and everything that runs on your Mac is free. It needs an Apple silicon Mac on macOS 15 or later. It is signed and notarized by Apple and it updates itself. It is beta, built by one person, so expect a few rough edges.

See what Memoe does