MCP ConfigsMay 13, 2026·3 min read

Osaurus — macOS AI Harness with Sandbox + MCP

Osaurus is a native macOS AI harness for agents, memory, and tools: run local/offline models, start a server/CLI, and expose tools via an MCP server.

MCP Hub
MCP Hub · Community
Agent ready

This asset can be read and installed directly by agents

TokRepo exposes a universal CLI command, install contract, metadata JSON, adapter-aware plan, and raw content links so agents can judge fit, risk, and next actions.

Native · 94/100Policy: allow
Agent surface
Any MCP/CLI agent
Kind
Mcp
Install
Brew|Dmg
Trust
Trust: Established
Entrypoint
brew install --cask osaurus
Universal CLI install command
npx tokrepo install f671a91f-6ee3-5d9e-a879-8f83ac23f94d
install contractmetadata JSONadapter planraw content
Intro

Osaurus is a native macOS AI harness for agents, memory, and tools—run offline local models or connect to cloud providers, then expose tools via an MCP server. The repo is GitHub-verified at 5,230★ and documents a brew cask install.

Best for: macOS users who want one harness for agents + memory + tools across local and cloud models

Works with: macOS Apple Silicon; local/offline models; OpenAI/Anthropic/Ollama-compatible APIs; MCP server mode

Setup time: 15–30 minutes

Key facts (verified)

  • README provides brew install --cask osaurus and osaurus ui/serve/status commands.
  • README notes sandboxed execution in an isolated Linux VM on macOS 26+ (Tahoe).
  • README lists MCP server config snippet: command: osaurus with args ["mcp"].
  • GitHub: 5,230 stars · 275 forks; pushed 2026-05-13 (GitHub API verified).

Main

If you are evaluating multiple models/providers, Osaurus helps because the harness (agents + memory + tools) stays constant while models change.

A good first week plan:

  • Install and run the UI + server.
  • Create one agent with a dedicated working folder.
  • Turn on the sandbox only for tasks that need shell execution, and keep workloads separated per agent.

Once stable, connect an MCP client so your other tools can reuse Osaurus-managed capabilities.

README excerpt (verbatim)

Screenshot 2026-03-19 at 3 42 04 PM

Osaurus

Own your AI.
Agents, memory, tools, and identity that live on your Mac. Built purely in Swift. Fully offline. Open source.

Release Downloads License Stars

Platform OpenAI API Anthropic API Ollama API MCP Server Foundation Models Hugging Face PRs Welcome

Download for Mac · Docs · Models · Discord · Twitter · Plugin Registry

Inference is all you need. Everything else can be owned by you.

FAQ

Q: Is Osaurus offline? A: README says it works fully offline with local models, and can also connect to cloud providers when desired.

Q: How do I install it? A: Install via brew install --cask osaurus or download the DMG from GitHub releases.

Q: How do I expose it via MCP? A: Add an MCP server entry with command: osaurus and args: ["mcp"] per README.

🙏

Source & Thanks

Source: https://github.com/osaurus-ai/osaurus > License: MIT > GitHub stars: 5,230 · forks: 275

Discussion

Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.

Related Assets

mcp-server-apple-events — macOS Reminders MCP

mcp-server-apple-events exposes macOS Reminders/Calendar actions to MCP clients via Apple Events, with an `npx` quick start and ready-to-copy configs.

MCP Configs
MCP Hub

Peekaboo — macOS Screenshots + Clicks (MCP)

Peekaboo is a macOS CLI (optional MCP) for screenshots and click automation; verified 3,986★ and installs via Homebrew or `npx -y @steipete/peekaboo`.

MCP Configs
MCP Hub

LitterBox — Self-Hosted Payload Sandbox (MCP)

LitterBox is a self-hosted payload-analysis sandbox: upload a sample, run static and EDR analysis, and review a Detection Score before it leaves the lab.

MCP Configs
MCP Hub

Apify MCP Server — 8,000+ Web Scrapers for Agents

Apify MCP Server connects agents to Apify Actors via a hosted endpoint (mcp.apify.com) or local run, turning thousands of web scrapers into callable tools.

MCP Configs
MCP Hub
◈Home🔍Search👤Me