As artificial intelligence models become more widely available and interchangeable, a growing number of startups are focusing on the software that sits on top of them—the layer that controls how users interact with these models. One notable newcomer in this space is Osaurus, an open-source LLM server designed exclusively for Apple devices. It allows users to switch between different local AI models, whether running directly on their machine or through the cloud, while keeping their files and tools securely stored on their own hardware.

Osaurus originated from an earlier project called Dinoki, a desktop AI companion that co-founder Terence Pae described as something like an “AI-powered Clippy.” Users of Dinoki questioned why they should pay for the app when they still had to cover the cost of tokens—the usage fees charged by AI companies for processing prompts and generating responses. This feedback prompted Pae to explore the possibility of running AI entirely on local machines. “You can do pretty much everything on your Mac locally, like browsing your files, accessing your browser, accessing your system configurations,” he explained. “I figured this would be a great way to position Osaurus as a personal AI for individuals.”

Pae began developing the tool in public as an open-source project, continuously adding features and fixing bugs along the way.

Today, Osaurus offers flexible connectivity with both locally hosted AI models and cloud-based providers like OpenAI and Anthropic. Users can freely choose which AI model they want to use, while keeping other aspects of the AI experience—such as the model’s memory, files, and tools—on their own hardware. Since different AI models have different strengths, this setup allows users to switch to the model that best suits their needs. This structure makes Osaurus what is known as a “harness”—a control layer that connects various AI models, tools, and workflows through a single interface, similar to tools like OpenClaw or Hermes. However, those tools are often aimed at developers comfortable with command-line interfaces, and some, like OpenClaw, can present security vulnerabilities. Osaurus, by contrast, offers an easy-to-use interface for consumers and addresses security concerns by running tasks in a hardware-isolated, virtual sandbox. This limits the AI’s reach, keeping the user’s computer and data safe.

Of course, running AI models locally is still in its early stages, as it requires significant computing resources and depends heavily on hardware. To run local models, a system needs at least 64 GB of RAM. For larger models like DeepSeek V4, Pae recommends systems with around 128 GB of RAM. Still, he believes the resource demands of local AI will decrease over time. “I can see the potential of it, because the intelligence per wattage—which is like the metric for local AI—has been going up significantly. It’s on its own curve of innovation. Last year, local AI could barely finish sentences, but today it can actually run tools, write code, access your browser, and order stuff from Amazon… it’s just getting better and better,” he said.

Osaurus currently supports models such as MiniMax M2.5, Gemma 4, Qwen3.6, GPT-OSS, Llama, DeepSeek V4, and others. It also works with Apple’s on-device foundation models and Liquid AI’s LFM family of on-device models. On the cloud side, it can connect to OpenAI, Anthropic, Gemini, xAI/Grok, Venice AI, OpenRouter, Ollama, and LM Studio. As a full MCP (Model Context Protocol) server, it allows any MCP-compatible client to access the user’s tools. Additionally, Osaurus ships with more than 20 native plugins for apps like Mail, Calendar, Vision, macOS Use, XLSX, PPTX, Browser, Music, Git, Filesystem, Search, Fetch, and more. Recently, it was updated to include voice capabilities as well.

Since the project launched nearly a year ago, it has been downloaded over 112,000 times, according to its website. The founders—including co-founder Sam Yoo—are currently participating in the New York-based startup accelerator Alliance. They are also exploring next steps, which could involve offering Osaurus to businesses in sectors like legal or healthcare, where running local LLMs could address privacy concerns. As local AI models become more powerful, the team believes they could reduce the demand for AI data centers. “We’re seeing this explosive growth in the AI space where [cloud AI providers] have to scale up using data centers and infrastructure, but we feel like people haven’t really seen the value of the local AI yet,” Pae said. “Instead of relying on the cloud, they can actually deploy a Mac Studio on-prem, and it should use substantially less power. You still have the capabilities of the cloud, but you will not be dependent on a data center to be able to run that AI.”