Osaurus brings local and cloud AI models to your Mac

As AI models become increasingly commoditized, startups are racing to build the software layer that sits on top of them. One interesting entrant in this space is Osaurus, an Apple-only open source LLM server that allows users to move between different native AI models, either locally or in the cloud, while keeping their files and tools all on their own devices.

Osaurus evolved from the idea of ​​a desktop AI companion, Dinoki, which Osaurus co-founder Terence Pye described as a kind of “AI-powered Clippy”. Dinoki customers asked him why they should buy the app if they still have to pay for tokens, which are the units of use that AI companies charge for processing claims and generating responses.

This has prompted Bai to think more deeply about running AI locally.

“That’s how Osaurus started,” Pai, who previously worked as a software engineer at Tesla and Netflix, told TechCrunch over a phone call. The idea is to try to get the AI ​​assistant running locally, he explained. “You can do almost everything on your Mac locally, like browsing your files, accessing your browser, accessing your system configurations. I thought this would be a great way to position Osaurus as a personal AI for individuals.”

Pai began building the tool publicly as an open source project, adding features and fixing bugs along the way.

Today, Osaurus can flexibly connect with AI models hosted locally or cloud providers like OpenAI and Anthropic. Users can freely choose which AI models they use, and keep other aspects of the AI ​​experience on their own devices, such as their own model memory, or their own files and tools.

Since different AI models have different strengths, the advantage of this system is that users can switch to the AI ​​model that best suits their needs.

Such an architecture makes Osaurus a so-called “belt” – a control layer that connects different AI models, tools and workflows through a single interface, similar to tools like OpenClaw or Hermes. However, the difference is that such tools are often aimed at developers who know their way around a terminal. Sometimes, as in the case of OpenClaw, they may present security issues and vulnerabilities that are cause for concern.

Meanwhile, Osaurus offers a user-friendly interface that consumers can use, and addresses security concerns by running things in a virtual sandbox isolated from the hardware. This limits the AI ​​to a certain range, keeping your computer and data safe.

Of course, the practice of running AI models on your device is still in its early days, as it is resource-intensive and hardware-dependent. To run local models, your system will need at least 64GB of RAM. To run larger models, like the DeepSeek v4, Pae recommends systems with around 128GB of RAM.

But Pai believes that domestic AI needs will decline in time.

“I can see the potential for this, because intelligence per wattage — which is like a metric for domestic AI — is going up significantly. It’s on its own innovation curve. Last year, domestic AI could barely finish sentences, but today it can run gadgets, write code, access your browser, order stuff from Amazon.” […] “It gets better and better,” he said.

Today’s Osaurus can run MiniMax M2.5, Gemma 4, Qwen3.6, GPT-OSS, Llama, DeepSeek V4, and other models. It also supports Apple’s core on-device models, Liquid AI’s LFM family of on-device models, and in 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 give any MCP-compliant client access to your tools as well. Additionally, it comes with 20+ native plugins for Mail, Calendar, Vision, macOS usage, XLSX, PPTX, Browser, Music, Git, Filesystem, Search, Fetch, and more.

Recently, Osaurus was updated to include audio capabilities as well.

Since the project was launched nearly a year ago, it has been downloaded 112,000 times, according to its website.

Currently, the founders of Osaurus (including co-founder Sam Yoo) participate in the New York-based startup accelerator alliance. They are also considering next steps, which could see Osaurus introduced to companies, such as those in the legal space or in healthcare, where running LLMs locally could address privacy concerns.

As homegrown AI models become more powerful, the team believes it could reduce demand for AI data centers.

“We are seeing this tremendous growth in artificial intelligence where [cloud AI providers] “We should be expanding with data centers and infrastructure, but we feel like people haven’t really seen the value of native AI yet,” Pai said. “Instead of relying on the cloud, they can actually deploy Mac Studio on-premises, and it should use a lot less power. You still have the capabilities of the cloud, but you’re not going to rely on the data center to be able to run that AI,” he added.