DOWNLOAD

One file. Two minutes.

macOS
READY
.dmg · Apple Silicon and Intel · macOS 13+

Alfred detects the right version himself. You download one file; onboarding handles the rest.

You're among the first 1,000. Alfred Wolf is yours — 3 months free, full access.
Windows
READY
.exe · same codebase

System requirements

Mit lokalem KI-Modell (volle Privatsphäre)
+macOS: Apple M1 oder neuer, 16 GB RAM, 15 GB Speicher
+Windows: NVIDIA-Grafikkarte empfohlen, 16 GB RAM, 15 GB Speicher
+Empfohlen: Apple M2/M3, 32 GB RAM
+Antwortzeiten lokal: Apple-Chips wenige Sekunden, Intel-Prozessoren 30–40 Sekunden pro Antwort (ohne Grafikbeschleunigung normal)
Mit Cloud-Anbindung (läuft auf jedem Rechner)
+macOS 13+, 8 GB RAM, 5 GB Speicher
+Windows 10/11, 8 GB RAM, 5 GB Speicher
+Zusätzlich: Claude Max / ChatGPT Plus Abo oder API-Schlüssel
Note
Alfred erkennt deine Hardware automatisch und wählt das passende Modell (Standard: Qwen 3.5, ~3,4 GB Download). Auf Intel-Rechnern dauern lokale Antworten 30–40 Sekunden — mit Cloud-Anbindung antwortet Alfred immer in Sekunden.

And then?

01
Open and confirm

Click the file, drag Alfred Wolf into Applications, launch it. The first time you confirm the start once — after that never again.

02
Onboarding

He looks at your machine, loads the fitting local models, and asks about you and your first project.

03
Set the trust level

The default is Level 1: he reads and suggests, changes nothing. You go higher when you're ready.

04
Let him work

From now on he runs alongside you. Every four hours he checks in, condenses, and reports what you need to know.

AT LEAST
What your machine needs
16 GB RAM · smooth from 32 GB
20 GB free for models and memory
Apple Silicon or x86_64

Alfred checks this himself on first launch. If the machine isn't enough, he says so and suggests running on a server.

ALTERNATIVE
On a server

For weak hardware, Alfred runs as a container on your own VPS. Access through the browser, data stays on your machine.

docker pull alfredwolf/alfred:latest