nativ
Local AI, native to your Mac. Chat, serve, monitor, and connect MLX models from one macOS app.
What is it?
What it is
Nativ is a native macOS workspace that lets users run AI models locally on Apple silicon. It bundles an `mlx-vlm` Python server, discovers compatible models in the Hugging Face cache, and exposes a polished SwiftUI interface for chatting, monitoring, and managing models.
Why it exists
It provides a privacy‑first, high‑performance local AI platform that eliminates the need for cloud inference, supports a wide range of models, and offers OpenAI‑ and Anthropic‑compatible APIs for developers and coding tools.
Who should use it
macOS developers, AI enthusiasts, and professionals seeking private, local LLM inference with OpenAI/Anthropic-compatible APIs and integration with coding tools.
Who should avoid it
Users without Apple silicon Macs, those requiring cross-platform support, or individuals uncomfortable with managing AI models and basic CLI/troubleshooting.
How it works
A quick walkthrough in plain English
How nativ works
Step 1 of 3
You interact with it
Open nativ, send a request, or connect it to your stack.
Features
Advantages
- Native macOS app optimized for Apple silicon
- Runs AI models locally without cloud dependency
- Polished SwiftUI interface with menu bar integration
- Supports multiple model types and coding agent integrations
- Detailed performance metrics and analytics
- Secure local API endpoints with optional API key protection
Disadvantages
- Requires macOS 26+ and Apple silicon
- Initial setup and model downloads need network access
- Limited to models compatible with MLX runtime
- No support for audio-only or image-generation-only models yet
- First-time build can be time-consuming due to Python bundle assembly
Installation
native
Download the latest DMG from GitHub Releases, drag Nativ to Applications, and launch it.
FAQ
What are the system requirements to run Nativ?
You need a Mac with Apple silicon, macOS 26 or newer, and enough unified memory for the model you choose. A Hugging Face token is optional for downloading gated models.
How do I get started with Nativ after downloading the release?
Drag Nativ to Applications, launch it, then choose or download a model, optionally generate an API key for the server, open the Models tab to manage models, and start chatting or view analytics.
Can I use Nativ as a local API server for other tools?
Yes. By default the server runs on http://127.0.0.1:8080 and provides OpenAI‑compatible endpoints (chat, responses, image, audio, models) and Anthropic‑compatible /v1/messages, plus health, metrics, and cache controls. You can change the port in the Developer page and authenticate with a Bearer token if you set an API key.
How does Nativ discover and manage models?
Nativ scans your Hugging Face cache (respecting HF_HUB_CACHE and HF_HOME) for compatible MLX models, lets you browse and download new models with memory fit warnings, inspect capabilities, switch models, or remove old ones from the Models library.
What kind of performance data does Nativ provide?
The Dashboard shows request volume, token usage, time to first token, decode speed, model performance, and recent activity, allowing you to monitor inference metrics over time.
What are the steps to build Nativ from source?
Install Xcode with the macOS 26 SDK, xcodegen, and Python 3. Then run `brew install xcodegen`, `make xcode-generate`, `make xcode-build`, and finally open the generated .app in build/XcodeDerivedData/Build/Products/Debug.
Featured in Videos
YouTube tutorials and walkthroughs for nativ
Alternatives
Similar projects ranked by category, topics, and text overlap.
