Mac users with Apple Silicon (macOS 15.5+) who want a personal agent app with local models; knowing what an LLM and an API key are helps.

What can you build with osaurus?

Run a private chat assistant on your Mac using a local MLX model, fully offline. Create a coding agent with a working folder and sandbox that plans and executes tasks with a todo list. Give an agent a persistent browser session to navigate sites and fill forms with your approval. Point existing OpenAI- or Anthropic-compatible tools at a local server provided by Osaurus. Generate or edit images offline with a local image model, right inside a chat.

osaurus-ai /

osaurus

A native macOS app for running AI agents with local or cloud models

8.1k stars
51 stars since 3 Oct
Swift
Updated today
Beginner friendly
Models & local AI

Open on GitHub (opens in a new tab) Website (opens in a new tab)

Osaurus is a layer that sits between you and any AI model. The model does the thinking. Osaurus keeps the rest on your Mac: agent prompts, memory, tools and identity. You can swap models without losing any of it.

Each agent has its own prompts, memory and theme. Give an agent a working folder and it gets file, search and git tools. Turn on the sandbox and it can run shell commands inside an isolated Linux VM (macOS 26+) or a Seatbelt sandbox on older systems. The model writes a todo list, works through it and ends with a summary.

It runs local models through MLX and can also connect to cloud providers. The README says it exposes OpenAI-, Anthropic- and Ollama-compatible APIs and works as an MCP server. A built-in Orchestrator can hand tasks to other agents or models as subagents. Compared with Electron-based chat apps, it is written natively in Swift.

What it does

Agent loop with sandbox
Agents get file, search and git tools in a working folder. With the sandbox on, they run code in an isolated Linux VM, with a Seatbelt fallback on older macOS.
Layered memory
Memory has identity, pinned facts and per-session episodes. Conversations are distilled once at session end, and most turns add about 800 tokens or fewer.
Privacy filter for cloud sends
An on-device classifier plus regex rules detect names, emails, keys and similar data before a cloud request. You review a scrubbed preview, and the send is blocked if anything leaks.
Identity, relay and encryption
Agents get secp256k1 addresses and scoped access keys. They can be reached through a relay, and agent-to-agent traffic is end-to-end encrypted.
Local models and image generation
Runs models like Gemma, Qwen, GPT-OSS and Llama via MLX, plus Apple Foundation Models on macOS 26+. Local image models such as FLUX can generate and edit pictures offline.

What you could build

Run a private chat assistant on your Mac using a local MLX model, fully offline.
Create a coding agent with a working folder and sandbox that plans and executes tasks with a todo list.
Give an agent a persistent browser session to navigate sites and fill forms with your approval.
Point existing OpenAI- or Anthropic-compatible tools at a local server provided by Osaurus.
Generate or edit images offline with a local image model, right inside a chat.

Get started

Install with Homebrew or download the .dmg from Releases, then open Osaurus from Spotlight and pick a model during onboarding. The ~4 GB build bundles a model so you can chat offline right away.

brew install --cask osaurus

osaurus ui

osaurus serve

Commands are copied from the project's README. Read it on GitHub for the full setup.