local-studio

Why we included this project

Keeping several LLM backends running on your own hardware usually means juggling a different CLI for each one. Local Studio puts vLLM, SGLang, llama.cpp, and exllamav3 behind one desktop control panel with a shared OpenAI-compatible proxy, so launching and evicting models, watching VRAM, and chatting with any of them all happen in the same place. It ships as a signed macOS app for Apple Silicon; a Bun/Hono controller handles model lifecycle, downloads, and logs, while the Next.js/Electron frontend adds a chat surface and an agent workbench. You can point the frontend at a remote controller on a GPU host, which makes it a useful laptop client for teams that want the same status and usage views no matter where the backend runs. Think of it as a management layer over inference servers rather than a replacement for them.