Osaurus brings local and cloud-based AI models to your Mac

As AI models become more and more advanced, developers are racing to create software that stays on top of them. One interesting entry into this space is Osaurus.

Osaurus was based on the concept of a desktop AI companionDinoki, co-founder of Osaurus Terence Pa is described as a version of “AI-powered Clippy.” Dinoki’s customers asked him why they should buy the app if they have to pay for tokens – the units used by the AI ​​companies to charge for processing the requests and answers.

This led Pae to think seriously about running AI locally.

“This is how Osaurus started,” Pae, a former software engineer at Tesla and Netflix, told TechCrunch by phone. The idea, he explained, was to try running an AI assistant locally. “You can do everything on your Mac, like browse your files, access your browser, access your system settings.”

Pae began to make a tool in public like open workadding features and fixing bugs along the way.

Today, Osaurus they can easily connect with AI models that live on-premises or with cloud providers like OpenAI and Anthropic. Users can freely choose which AI models they are using, and store some AI features on their devices, such as their own models’ memory, or their own files and devices.

Since different types of AI have different strengths, the advantage of this system is that users can choose the type of AI that best suits their needs.

This design creates Osaurus’ so-called “harness” – a control unit that connects different types of AI, tools, and workflows through a single interface, similar to tools such as OpenClaw or Hermes. However, the difference is that such tools are often aimed at developers who know their way around a terminal. And in some cases, like with OpenClaw, they can cause security problems and loopholes to complain about.

Osaurus, meanwhile, has an easy-to-use interface that consumers can use, and addresses security concerns by running things in a self-contained sandbox. This slows down the AI ​​to some extent, keeping your computer and data safe.

Of course, the process of using AI models on your machine is still in its early days, as it is very complex and hardware dependent. To use the local samples, your system will need at least 64 GB of RAM. For running larger versions, such as DeepSeek v4, Pae recommends systems with at least 128 GB of RAM.

But Pae believes the need for local AI will decline over time.

“I see its potential, because each intelligence — which is like a local AI metric — has been going up a lot. It’s on its own technology path.” Last year, local AI could not complete sentences, but today it can use tools, write code, access your browser, and order products from Amazon (…) it’s just getting better,” he said.

Osaurus today can run MiniMax M2.5, Gemma 4, Qwen3.6, GPT-OSS, Llama, DeepSeek V4, and other models. It also supports the original models on Apple devices, the LFM family of Liquid AI devices, and on the cloud, it can connect to OpenAI, Anthropic, Gemini, xAI/Grok, Venice AI, OpenRouter, Ollama, and LM Studio.

As a full MCP (Model Context Protocol) server, you can also give any MCP-compliant client access to your devices. In addition, it ships with more than 20 plugins for Mail, Calendar, Vision, Use MacOS, XLSX, PPTX, Browser, Music, Git, Filesystem, Search, Fetch, and more.

Recently, Osaurus has been updated to include audio capabilities as well.

Since the project began nearly a year ago, it has been downloaded north of 112,000 times, according to website.

Meanwhile, the founders of Osaurus (which includes co-founder Sam Yoo) are participating in the New York-based startup accelerator Alliance. He’s also thinking about the next steps, which could see Osaurus offered to businesses, such as those in the legal or medical fields, where running local LLMs can address privacy concerns.

As the power of conventional AI grows, the team believes it can reduce the need for data center AI.

“We’re seeing a lot of growth in the AI ​​space as (cloud AI providers) have to expand their use of data centers and infrastructure, but we feel like people still haven’t seen the benefits of native AI,” Pae said. “Instead of relying on the cloud, they can deploy Mac Studio on-prem, and it should use very little power. You have the power of the cloud, but you won’t rely on a data center to run AI,” he added.