LocalModel Bridge

Privacy Commitment

The protection of your personal data and AI conversations is an absolute priority. LocalModel Bridge is designed to respect your privacy fully. No prompts, processed text, or model outputs are collected, analyzed, sold, or shared for advertising or commercial purposes.

100% Offline and Local Processing

LocalModel Bridge operates **fully locally**. All inference operations (text generation, summarization, chatting) run in-process on your Mac using Apple's MLX Swift framework on the Apple Silicon Unified Memory. Your prompts never traverse any third-party cloud servers.

App Sandbox & Network Permissions

The application runs within a secure macOS App Sandbox. Its network activities are strictly limited to the following features:

• Hugging Face DownloadsOutgoing client network access is used solely to query and download public model weight files from the Hugging Face Hub, triggered only on your explicit request.
• Local API ServerServer network access is used solely to host the local OpenAI-compatible HTTP server (`localhost`), permitting other development apps on your Mac (e.g. Xcode, Cursor) to interact with your models. This local server is never exposed to the outside network.

Model and Data Storage

Models downloaded from Hugging Face are stored locally in the application's containerized cache folder. Your model profiles and security bookmarks (required to access local weight folders securely) are saved on your device using SwiftData.

Trackers and Advertising

The application contains no analytics SDKs, trackers, or advertising frameworks. We do not monitor your usage patterns, model selections, or prompt content.