Checking status…

MacLustr AI — Intelligence, running locally.

High-performance AI inference powered by Apple Silicon and MLX.

  • OpenAI-compatible
  • SSE streaming
  • Tool calling
  • Validated JSON
  • On-device inference
One request away

If you speak OpenAI, you already speak MacLustr.

Keep your SDK, your message format and your streaming loop. Change the base URL, use a MacLustr key, and call the model served on our hardware.

Base URL
https://llm.spboucher.ai/v1
Model
maclustr-qwen3
Auth
Authorization: Bearer mlx_sk_…
Platform

Inference infrastructure, not a chat widget.

Everything below is implemented in the gateway today and described precisely in the documentation.

Architecture

A short, observable path from request to token.

Requests are authenticated, metered and queued by priority before they reach the MLX engine. The engine and the gateway run on the same Apple Silicon machine.

Request pathClient apps connect over HTTPS to a TLS gateway, which forwards to the API Gateway (authentication, rate limits, priority queue). The API Gateway feeds the MLX engine (continuous batching, KV prefix cache), which runs on the Apple Silicon GPU and unified memory. 01Client appsOpenAI SDKscURL · URLSessionYour backendHTTPS02HTTPS / TLS gatewayPublic endpointTLS termination03API GatewayAuthenticationRate limitsPriority queue04MLX engineContinuous batchingKV prefix cacheMetal05Apple SiliconGPU (Metal)Unified memorygateway processengine process · same machine Request pathClient apps connect over HTTPS to a TLS gateway, which forwards to the API Gateway (authentication, rate limits, priority queue). The API Gateway feeds the MLX engine (continuous batching, KV prefix cache), which runs on the Apple Silicon GPU and unified memory. 01Client appsOpenAI SDKscURL · URLSessionYour backendHTTPS02HTTPS / TLS gatewayPublic endpointTLS termination03API GatewayAuthenticationRate limitsPriority queue04MLX engineContinuous batchingKV prefix cacheMetal05Apple SiliconGPU (Metal)Unified memory

Start with one request.

Read the five-minute guide, or try the model against the real API in the Playground.