MacPaw is partnering with Liquid AI to build a local inference stack for its Eney assistant and eventually open the same runtime to third-party developers on its SetApp store, which has over 150,000 paying users. The Ukraine-based developer unveiled Eney last year as a cloud-connected assistant; the new work with Liquid AI produces a locally hosted version powered by an on-device inference engine called Elix, along with a local memory system.
The pitch is straightforward: run the assistant on the Mac itself, get privacy and offline capability by default, and give SetApp's subscriber base a reason to keep paying as AI features become the differentiator across desktop software. MacPaw is pairing the technical shift with a commercial one, testing credit-based pricing where users draw down a pool of AI operations weighted by task complexity rather than a flat monthly fee.
Liquid AI's argument for why a third-party on-device engine can compete with Apple's own local models rests on architecture choice. The startup designs models around the target hardware first, rather than compressing a general-purpose model to fit.
“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.”— Ramin Hasani, Liquid AI co-founder and CEO
Key facts
- 01MacPaw's SetApp store has over 150,000 paying subscribers, the base that will get access to on-device AI apps.
- 02Liquid AI is building an on-device inference engine called Elix, plus a local memory system, for MacPaw's Eney assistant.
- 03MacPaw is testing credit-based pricing that meters AI operations by task complexity rather than flat subscription tiers.
- 04The platform will also route to cloud models from Google, positioning SetApp as a one-stop AI runtime for third-party developers.
Apple already ships local models to developers through its own frameworks, and any Mac-native AI stack that isn't Apple's has to justify the integration cost. Liquid AI CEO Ramin Hasani argues that competing on raw capability per watt is where a specialist wins, and that MacPaw's willingness to let a partner own the runtime is the opening.
MacPaw CEO Oleksandr Kosovan frames the offline angle as a user-facing feature, not just a privacy talking point. Agentic workflows that can execute without a network round-trip open up scenarios — travel, secure environments, spotty connectivity — where cloud-only assistants stall.
“locally hosted AI models will also give users the ability to run assistants and agentic workflows offline.”— Oleksandr Kosovan, MacPaw CEO
SetApp's role in the plan is the distribution layer. Once the local inference architecture is stable, MacPaw wants to expose it to the developers already shipping apps on the store, so a third-party Mac utility can bake in AI features without standing up its own model pipeline. Kosovan said the platform will also broker access to cloud models from providers such as Google, positioning SetApp as a one-stop shop rather than a local-only silo.
The customization layer is where Liquid AI is trying to build durable lock-in. Hasani said the roadmap includes a stack that lets models adapt to individual user data over time.
“We are also building a customization stack around models. This means that with user input, the models can use the data and improve. We want our models to be adaptable and become more intelligent over time.”— Ramin Hasani, Liquid AI co-founder and CEO
For MacPaw, the timing matters. SetApp's subscription model depends on continuously justifying the bundle, and generative AI features have quickly become the category most subscribers evaluate first. Rolling out credit-based AI pricing on top of a flat subscription is an experiment in extracting incremental revenue from users who want heavier AI usage, without breaking the bundle for casual users.
For Liquid AI, the SetApp deal is a distribution beachhead. The startup has been building models designed for edge deployment, and landing an app store with a six-figure paid user base is a faster path to real-world inference volume than winning individual developer relationships one at a time. If Elix performs on Mac hardware, the same runtime is a candidate for Windows and mobile app stores that face the same offline-and-privacy pressures.
The open question is whether developers on SetApp actually want a non-Apple local runtime badly enough to integrate against it. Apple's frameworks are the path of least resistance on Mac, and any third-party stack has to demonstrate meaningfully better output quality, latency, or per-app customization to earn the switch. Neither MacPaw nor Liquid AI has published benchmarks against Apple's local models yet.
MacPaw is making a bet that owning the AI runtime — not just the storefront — is what turns SetApp from a utility bundle into an AI platform. If credit-based pricing sticks and developers ship apps built on Elix, the store becomes a rare independent on-device AI distribution channel on the Mac, competing directly with Apple's own developer tooling. If it doesn't, MacPaw is still shipping a differentiated version of Eney and a Ukraine-based counterweight to the assumption that every consumer AI product has to route through a US hyperscaler.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
