Overview
Nativ is a native macOS workspace for running open AI models locally on Apple silicon. It bundles an mlx-vlm server, discovers MLX models already in your Hugging Face cache, and wraps everything in a SwiftUI app you can use as a private chat client, a model manager, a performance dashboard, or an OpenAI- and Anthropic-compatible inference server on localhost. Chat with text and vision models, generate and edit images, connect coding agents like Claude Code, Codex, OpenCode, Aider, and Goose to a local endpoint, attach MCP servers, schedule routines, and dictate into any app with the bundled audio extension. Every token is generated on your machine — no accounts, no subscriptions, no cloud.
Minimum macOS: 26.0 (Tahoe)
Architecture: Apple Silicon
Key Features
- Runs MLX models fully offline on Apple silicon
- Streaming chat with vision, image attachments, and reasoning output
- Local image generation and editing with MLX image models
- OpenAI- and Anthropic-compatible endpoints on localhost
- Connect Claude Code, Codex, OpenCode, Aider, Goose, and Crush to local models
- Model library with Hugging Face downloads and memory fit warnings
- Performance analytics and live CPU, GPU, memory, and thermal monitoring
- MCP server support with per-tool consent
- Global voice dictation and transcript history via the audio extension
- Routines for scheduling prompts to run on their own
- Artifacts browser with semantic search over your own files
- Menu bar controls for the server and loaded model
