---
title: Minne and Raycast AI — which one is for you?
description: An honest comparison. Raycast is a launcher with excellent AI built in; Minne is a single-purpose memory that writes where you type. Where they overlap, and how to choose.
---


This is not a fair fight in either direction. [Raycast](https://www.raycast.com)
is a launcher — one of the best pieces of Mac software of its generation —
that has grown a serious AI layer. Minne is a small, single-purpose tool: a
memory of your work that writes from it. They overlap on exactly one thing,
AI-assisted writing on a Mac, and they come at it from opposite ends.

The short version: **Raycast AI** gives you many models and many AI features
inside Raycast's own windows. **Minne** gives you one thing — drafts grounded
in a memory of your work — inside whatever window you were already in.

## What Raycast AI does well

From Raycast's own site, checked at the date in the footer:

- **Quick AI** — a hotkey away, anywhere, with web search and inline
  references for current answers.
- **AI Chat** — 50+ models from OpenAI, Anthropic, Google, Mistral and
  others in one interface, with file attachments and screen context.
- **AI Commands and Extensions** — natural language that becomes actions:
  rename files, check Linear tickets, run your tools.
- **Pro features including Screen Awareness and a Memory feature**, per
  Raycast's published feature list.
- Polished, actively developed, with iOS and Windows versions.

Pricing at time of checking: AI arrives with Raycast Pro at $8/month billed
annually ($10 monthly), with an Advanced AI add-on at a further $8/month for
the premium models. Data is stored locally by default, with optional
encrypted cloud sync.

## Where Minne differs

- **The context is your accumulated work, not your current screen.** Minne
  [keeps a wiki](/what-is-ai-memory) of the people, projects and decisions
  it has watched you work on — distilled every half hour, as plain markdown
  in `~/Minne` — and every draft consults it. The reply to Ingrid knows
  what you promised Ingrid.
- **It writes in your field, not in a panel you paste from.** Tap
  right-Option in the reply itself; press Insert and the draft lands where
  the caret was. [One key, everywhere you type](https://docs.minne.sh/guides/minne-key).
- **In your tone.** Minne learns how you write per app and per person, from
  your own sent writing, and your repeated corrections
  [become standing rules](https://docs.minne.sh/guides/instructions).
- **Free, MIT-licensed, no account.** The memory is files you own, from
  [code you can read](https://github.com/mandgie/minne). You bring the model:
  an API key, [Ollama locally](/local-ai-writing), or a subscription sign-in.

## Choose Raycast if…

- You want a launcher — that is the product, and it is excellent.
- You want many models, web-connected answers, and AI that *does things*
  (commands, extensions, automations).
- You want mature, supported software across Mac, iOS and Windows.

## Choose Minne if…

- Your bottleneck is specifically the writing you repeat all day — replies,
  [status updates](/status-updates), answers you half-know — and you want it
  grounded in what you actually did.
- You want the memory itself on disk, in markdown, yours.
- You are happy running early, open-source software.

They also simply coexist: nothing about a launcher conflicts with a memory.
If the memory half is the gap in your setup, Minne is
[a free download](https://github.com/mandgie/minne/releases/latest/download/Minne.dmg)
for macOS 14+. See also [Minne and Goldfish](/vs/goldfish) — the closest
comparison there is — [Minne and Elephas](/vs/elephas), and, if you came from
a screen-recording tool, [the Rewind alternative page](/rewind-alternative).