Modular MAX on a Mac
Modular's inference server; OpenAI-compatible on :8000 · AI / LLMs
MAX is Modular's inference platform: it serves models from Hugging Face behind an OpenAI-compatible API, so anything that speaks /v1/chat/completions can use it. Install the platform with pip install modular, or max[all] via pip/uv (max-all on conda/pixi); Modular recommends pixi for the most reliable environment and uv for the tightest Python interop. Then max serve --model-path listens on :8000; point the AI Administrator at it as a custom endpoint, or LibreChat and the Model Costs pane. Serving performs best on NVIDIA/AMD GPUs under Linux; on this Mac expect fewer compatible models and slower generation. MAX and the Mojo language are free under Modular's Community License; Modular's paid products are its hosted services; Modular Cloud (per token or GPU hour), Bring Your Own Cloud (per minute, in your VPC) and Enterprise (custom); none of which are needed to self-host here. The Mojo language itself lives under Runtimes, where it installs on its own.
Run Modular MAX with FrontierStack
Install or connect Modular MAX; then check its status, ports and certificate from FrontierStack. The same screen links to firewall checks, Malware Audit and backups where they apply.
Run it from your Mac.
FrontierStack installs, monitors and secures services on this Mac and on linked servers.
Download FrontierStack