Run Claude Code entirely on-device using an MLX-native server on Apple Silicon

Description and use

Claude Code now runs entirely on Apple Silicon devices using MLX, eliminating API keys, reducing latency, and enhancing privacy.

Problem

Developers need to run Claude Code locally without relying on external API keys, reducing latency and privacy concerns.

Outcome

Claude Code runs entirely on Apple Silicon devices using MLX, eliminating API keys and reducing latency.

Workflow

• 1. Install MLX on Apple Silicon.
• 2. Clone the local Claude Code repository (https://github.com/nicedreamzapp/claude-code-local).
• 3. Set up the MLX-native server following the repo instructions.
• 4. Run Claude Code locally, ensuring the model is loaded into MLX.
• 5. Verify that no external API keys are used and that execution occurs on-device.

Evidence

Run Claude Code entirely on-device using an MLX-native server on Apple Silicon to execute local AI models without API keys.

Source