What anz33 is

anz33 is a small program that runs quietly on your computer. AI apps talk to it, and it passes each question along to whichever AI engine on your machine will answer fastest. You install it once and mostly forget it exists.

The problem it solves

Say you have installed Ollama to run models locally. Then a friend recommends LM Studio, so you install that too. Then an app you like only works with llama.cpp. Now you have three AI engines on one computer, each with its own address, each holding different models — and every app you own is hard-wired to exactly one of them.

anz33 sits in front of all of them and offers a single address:

http://127.0.0.1:3300/v1

Point any app at that address and it can reach everything. anz33 decides which engine actually answers, based on which one is genuinely fastest on your hardware and whether you are on battery.

127.0.0.1 means "this computer". It is an address that never leaves the machine — a bit like writing a note and leaving it on your own desk. Nothing in that address is reachable from the internet.

What it does for you

One address, every engine

Ollama, LM Studio, llama.cpp and MLX all become one list of models. Apps stop caring which one you happen to be running.

Picks the fastest

anz33 times your engines with a real request and uses the measurements — not a guess, and not a benchmark from somebody else's computer.

Knows about your battery

On battery it prefers the engine that will not flatten you by lunchtime. Plugged in, it prefers whatever is quickest.

Nothing leaves by default

Cloud engines are off until you switch them on, and even then you can write rules about what may never leave this machine.

What you need

A Mac, a Windows PC or a Linux machine. It runs on Apple Silicon and Intel, on 64-bit and ARM.
At least one local AI engine installed — most people start with Ollama — or a cloud key if you would rather start there.
Nothing else. anz33 ships as one self-contained file; there is no Python to install and no account to create.

New here? Read these four

Install anz33

Pick the download for your computer and get it running. Five minutes.
Your first request

Prove it works by asking it something, and see which engine answered.
Connect your apps

Change one setting in the apps you already use so they go through anz33.
Make it fast

Run one command so anz33 knows how quick each engine really is here.

A note on the words

Some of these come up constantly, so here they are once:

Model	The AI itself — the thing that writes the answer. "Llama 3", "Mistral" and "Claude" are models.
Engine	The program that runs a model on your machine. Ollama and LM Studio are engines. One engine can hold many models.
Router	anz33. It does not answer anything itself; it chooses who does.
Local	On your computer. Nothing sent anywhere.
Cloud / remote	Someone else's computer, over the internet. Off unless you turn it on.
Token	Roughly three-quarters of a word. Speed is measured in tokens per second.
Daemon	A program that runs in the background with no window. anz33 is one.