Show HN: Strim – a Swift/MLX model server for coding agents on Apple Silicon

Category: ai-ml

Tags: apple-silicon, llm-inference, mlx, local-first, openai-compatible, coding-agents

Score: 6.3/10 (Innovation: 6, Technical: 7, Documentation: 7, Utility: 5)

Strim is a Swift/MLX-based local inference server optimized for coding agents on Apple Silicon, serving a Qwen3 27B model through an Anthropic/OpenAI-compatible HTTP API. It combines a single binary with a menu-bar app and claims meaningful throughput gains over other MLX servers on decode and shared-prefix agent workloads, while keeping data on-device.

Target audience: macOS developers running local coding agents (backend devs, AI engineers) with Apple Silicon and privacy constraints

Repository: https://github.com/megasoft1978/strim-engine · Swift · MIT

View on Hacker News