AI BriefingMac Coding Agent mlx-code Released
Key point
A lightweight, Mac-exclusive coding agent built on Apple's MLX framework has been released.
Details
mlx-code is a lightweight coding agent optimized for the Apple Silicon environment. Leveraging Apple's MLX framework, it supports fast local inference and provides prompt caching along with powerful tool-calling capabilities.
Rather than a complex UI, this tool aims for a terminal-centric interface, focusing on building a fast feedback loop that users can control by leveraging Unix primitives.
Key Components and Features:
- LLM Server and Harness:
main.pyloads quantized models and provides an OpenAI-compatible endpoint, whilepie.pyhandles agentic functionality. - Multi-Provider Support: Compatible not only with local MLX models but also with various APIs such as Claude, Gemini, DeepSeek, and OpenAI.
- Prompt Caching: Automatically saves and reuses the KV cache on disk to improve efficiency.
- CLI Tools: Intuitive control is possible via the
mc(local agent),me(harness), andmd(log viewer) commands.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Underdog Releases 'Husky', a Model-Specific Inference Engine Up to 4.5x Faster than MLX
Jared Palmer Releases 'Kev', a Lightweight Decision-Making Model Based on Qwen3.5