mlx-code

TL;DR

A local coding assistant for Apple Silicon that uses a swarm of small subagents to stop your LLM from getting confused by massive codebases.

Who is this actually for?

Mac-based developers who want to stop burning money on OpenAI API credits and value their privacy enough to run everything locally.

The Good

• Uses MLX for native performance on Mac hardware, so it actually feels fast.
• Subagenting prevents the context rot where the AI forgets the beginning of the file by the time it reaches the end.

The Catch (Potential Downsides)

If you aren't rocking an M2 or M3 Max with at least 64GB of unified memory, you will likely find it sluggish. It is also useless if you are on Windows or Linux.