MacPaw teams up with Liquid AI to run its assistant directly on your Mac
The company behind SetApp wants its AI tools to work offline, keep your data on your device, and eventually let developers plug into the same system.
Key points
- MacPaw, a Ukraine-based app developer, has partnered with Liquid AI to build an on-device AI inference system called Elix for its assistant, Eney.
- Liquid AI designs its models around specific hardware, so they run directly on the device without sending data to a remote server.
- SetApp, MacPaw's subscription app store with over 150,000 paying users, is preparing a credit-based pricing tier for AI features.
- MacPaw plans to open the on-device system to third-party developers once the architecture is finalised.
- Users will be able to run AI assistants and multi-step automated workflows, called agentic workflows, without an internet connection.
Most AI assistants send your words and questions to a distant server, crunch through them in the cloud, and ship the answer back. MacPaw wants to skip that round trip entirely.
The Ukrainian company behind SetApp, a subscription bundle of Mac and iPhone apps, has partnered with Boston-based Liquid AI to build a system that runs artificial intelligence directly on your device. The project has two parts: an on-device inference engine, the software that runs AI calculations locally rather than in the cloud, called Elix, and a local memory system that lets the AI remember things about you without storing that information anywhere else.
What does "on-device AI" actually mean for you?
In short: your data stays on your Mac. Nothing gets sent to a server.
Ramin Hasani, Liquid AI's co-founder and chief executive, told TechCrunch that his team picks a model architecture tailored to the specific chip inside the device before any training even begins. That tight fit between software and hardware is what lets a smaller model punch above its weight. "That allows us to really have the most efficient version of intelligence that runs directly on the device, with benefits like privacy and security," he said.
MacPaw's chief executive Oleksandr Kosovan added that running locally means the assistant keeps working even when you have no internet connection, including the kind of multi-step automated tasks, known as agentic workflows, where the AI books a meeting or drafts a document on your behalf.
Apple already gives developers their own local models through Apple Intelligence. Liquid AI's pitch is that its models focus on specific performance capabilities and can also adapt over time. "With user input, the models can use the data and improve," Hasani said. The model learns from you without that learning leaving your machine.
What changes for SetApp users?
SetApp already charges a flat monthly subscription. Now MacPaw is testing a credit-based layer on top, where each AI task costs a certain number of credits depending on how complex it is, similar to how some photo-editing apps charge per export. The store currently has over 150,000 paying subscribers.
Once the on-device system is solid, MacPaw plans to open it up to outside developers, so any app on SetApp could plug into local AI without building the infrastructure from scratch. The platform will also connect to cloud models from providers like Google for tasks that need more horsepower.
Should developers pay attention?
Yes, especially those building Mac apps where privacy is a selling point, such as health trackers, legal tools, or personal finance apps.
For ordinary users, the headline benefit is simple: an AI assistant that cannot leak your data because it never sends it anywhere. Watch for Eney to appear in the SetApp catalogue once the Elix system is ready.
Common questions
Is this the same as Apple Intelligence?
No. Apple Intelligence is Apple's own on-device AI layer built into macOS and iOS. Elix is a separate system built by Liquid AI specifically for MacPaw's apps, designed to be opened up to third-party SetApp developers as well.
Will this cost extra on SetApp?
MacPaw is testing credit-based pricing for AI features, so some tasks may draw from a credit balance rather than being bundled into the flat subscription. Exact pricing has not been announced yet.
Do I need a new Mac to use it?
MacPaw has not published minimum hardware requirements yet. On-device AI generally runs best on Apple Silicon chips (the M-series processors in Macs made from late 2020 onwards), but the company has not confirmed which devices will be supported.
