Install faxl
Get the code, install it, install your key. If you do not have a key yet, get one — it is free.
This page covers one Mac. To run faxl across several machines, see faxl on a cluster.
The release
Get the wheels
The Mac release is on GitHub: two Python wheels (binaries), the README,
and an AGENTS.md so your agent can drive it over MCP.
Without a licence key the proxy still serves every request, it just doesn't cache. The console shows the countdown to expiry.
Apple silicon
Install
You need a Mac with an M-series chip. In Terminal:
git clone https://github.com/Gamakon/faxl-mac-community.git
cd faxl-mac-community
./upgrade_install.sh
The script installs everything faxl needs (Python 3.12 or later and ffmpeg through Homebrew if they are missing) and starts faxl. Run it again in the same folder to upgrade: it fetches the latest release, stops the running faxl and starts the new one, keeping your licence, cache, models and settings. The README also has the steps by hand.
- Open the console at
http://127.0.0.1:8767/console. - Paste your licence key into the Licence panel, then restart the proxy. The key is read once at startup and verifies offline: no phone-home, no licence server.
- Pick a model. The first load downloads it from HuggingFace; models over 30 GB take several minutes.
- Point any OpenAI-compatible client at
http://127.0.0.1:8767/v1.
Full configuration, supported models and troubleshooting are in the README.
Other platforms
NVIDIA
The NVIDIA/vLLM build is not yet released. If you want it, tell us what you run and we will come back to you when it is ready — andrew@gamakon.ai.