Osaurus

About Osaurus

Osaurus is a free, MIT-licensed, native Swift app for Apple Silicon that runs open models entirely on your Mac and adds hosted models only when a task genuinely needs them — with one shared memory carried across both, so switching from a local model to ChatGPT, Claude, or Gemini does not reset the context you have built up. That shared-memory design is the substantive differentiator over running Ollama or LM Studio directly, where each frontend keeps its own history. Setup is deliberately minimal: open the .dmg and launch with no account, then connect Ollama, MLX, or LM Studio in one click to get a model running on your chip, fully offline. Inference happens on the M-series chip, so there are no server round trips, and with Wi-Fi off it still works. Beyond chat it reads files, runs tools, and executes work — drop in a folder, assign a task, and it can read, write, and execute in your actual environment. It also builds autonomous agents rather than only chat sessions: voice control, folder watchers, browser plugins, and parallel jobs let local models keep working unattended. Privacy is absolute by design — no telemetry, no training on your data, no anonymous analytics, and nothing leaves the machine unless you explicitly send it. Installation is a download or brew install osaurus, requiring macOS 15.5+ on Apple Silicon. The project publishes blog posts, docs, skills, themes, and a brand kit, is built in public under MIT, and is operated by Osaurus, Inc.

Key Features

Osaurus Pros & Cons

Who Is Osaurus Best For?

Tags