MacPaw is going local with AI. The Ukraine-based developer behind CleanMyMac and the SetApp store has partnered with Liquid AI to run AI models directly on users’ devices, and it plans to hand that stack to third-party developers, according to TechCrunch AI. The move sets up SetApp as a hub for AI apps that don’t need to phone the cloud for every request.
Here’s what the two companies are building together.
What’s happening
MacPaw is developing a locally hosted version of Eney, the AI assistant it unveiled last year. To power it, the company brought in Liquid AI to build two pieces:
- Elix, an on-device inference system that runs models directly on the hardware.
- A local memory system so assistants can retain context without sending data off the device.
MacPaw CEO Oleksandr Kosovan told TechCrunch AI that local models will let users run assistants and agentic workflows offline. Liquid AI co-founder and CEO Ramin Hasani explained the approach this way: “Before training our models, we select an architecture that is different and tailored to the hardware. That allows us to really have the most efficient version of intelligence that runs directly on the device, with benefits like privacy and security.”
Why it matters
On-device inference is one of the biggest shifts in AI right now. Running models locally means lower latency, no per-query cloud bills, and data that never leaves the user’s machine. For privacy-conscious users and developers watching their API spend, that’s a real draw.
What stands out here is the developer angle. Once MacPaw locks in the local architecture with Liquid AI, it wants to open the tech to developers building for SetApp, which already has over 150,000 paying subscribers. That turns a subscription app store into a distribution channel for AI-native apps, with the heavy lifting of local inference handled for them.
The Apple question
Apple already gives developers its own on-device models through its Foundation Models framework. So why would developers reach for Liquid AI instead?
Hasani’s answer is performance and customization. He says Liquid AI’s models focus on specific capabilities rather than being general-purpose, and the company is building a customization layer on top. “With user input, the models can use the data and improve. We want our models to be adaptable and become more intelligent over time,” he told TechCrunch AI. That pitch, models that adapt to the individual user over time, is the differentiator against Apple’s more standardized offering.
The business model
MacPaw isn’t just shipping tech. It’s rethinking how AI apps get paid for. The company is testing credit-based pricing on SetApp, where users get a pool of credits and each AI operation costs a certain amount depending on how complex the task is.
Kosovan also said the platform will offer access to cloud models from providers like Google, positioning SetApp as a one-stop shop where developers can mix local and cloud inference under one roof. That hybrid approach is smart. Some tasks run fine on-device, others need cloud horsepower. Giving developers both in one place lowers the friction of building AI features.
What to watch next
A few things worth tracking as this rolls out:
- Timing. MacPaw hasn’t said when the developer tools or the local Eney will ship. The local architecture has to be locked in first.
- Credit pricing. How MacPaw prices AI operations will signal whether credit-based models can work for a broad app store, not just single products.
- The Liquid AI bet. This is a notable commercial win for Liquid AI, the MIT spinout building efficient, hardware-tailored models. A real deployment across a 150,000-user store is a strong proof point.
The bigger story is that on-device AI is moving from research demos to shipping products, and app stores are starting to compete on it. MacPaw is betting that privacy, offline capability, and developer-friendly tooling can carve out space next to Apple’s own stack. You can read the full details at the original source.