VibeHunt
Back to browse

Osaurus

LLM server built on MLX

Visit

Osaurus provides a native macOS harness for running large‑language‑model agents entirely on‑device. It leverages the MLX framework to execute any compatible model at native Apple‑Silicon speed, keeping inference and all associated data local unless the user explicitly opts for cloud services. The platform offers persistent memory, sandboxed code execution, and a cryptographic identity that remain on the machine, ensuring privacy and control over keys and history.

Developers can integrate any model without rewriting code, swapping between local and cloud providers as needed. The Swift‑based implementation has no external runtime dependencies and runs on macOS 15.5 or newer, making it suitable for building autonomous agents that remember prior interactions, invoke real tools, and perform tasks without leaving the host system.

The project is open source under the MIT license, encouraging community contributions and custom extensions. It targets users who want a local‑first, privacy‑preserving AI environment on macOS, offering a single interface for model management, memory handling, and secure execution.

Reviews

Sign in to leave a review.

Loading reviews…

Similar apps