MacPaw Declares War on Cloud AI with Private Local Stack

Ukraine-based MacPaw is launching an assault on cloud-powered AI models, partnering with Liquid AI to build a fully local AI stack for macOS. According to the official announcement, the stack will first power MacPaw's upcoming AI assistant, Eney, before being offered to developers building for its Setapp marketplace. TechCrunch first reported the partnership.

The plan moves the core processing for AI tasks from centralized servers directly onto the user's Mac. This is a direct challenge to cloud-reliant services, pitching privacy, offline functionality, and speed as primary advantages. The joint architecture hinges on three technical layers: Liquid Foundation Models (LFMs) optimized for macOS, MacPaw's on-device inference system Elix, and a persistent memory layer called Mnemos.

Building a Private AI Assistant That Never Phones Home

The first product to deploy this stack will be MacPaw's own Eney, an AI assistant for macOS initially unveiled last year. The partnership aims to deliver results "later this year," transforming Eney from a concept into a locally-hosted tool.

"This combination of on-device intelligence, persistent memory, and native macOS task execution in a single product has not existed on the Mac before," MacPaw stated. For users, this translates to AI that can manage tasks without an internet connection and theoretically never sends personal data to a remote server.

"We believe intelligence should live where people work: private by design, fast by default, and be able to reach the cloud when that's the better tool," said MacPaw CEO Oleksandr Kosovan.

This approach directly confronts privacy concerns that have plagued cloud-based AI, such as data usage policies and potential exposure from breaches. It leverages MacPaw's deep macOS engineering history, which spans nearly two decades, to build what Kosovan calls "the AI stack for the Mac."

From In-House Tool to Developer Platform

The bigger play isn't Eney itself, but the infrastructure beneath it. Once the local processing architecture with Liquid AI is solidified, MacPaw intends to make the entire tech stack available to third-party developers on its platform.

Setapp, MacPaw's subscription-based app store, has over 150,000 paying users. By offering developers the same on-device inference and memory systems it uses for Eney, MacPaw aims to turn Setapp into a hub for privacy-focused AI applications. Kosovan said the platform will also provide access to cloud models from companies like Google, offering a hybrid, "one-stop shop" for developers.

Platform Strategy: Curation meets capability. This positions Setapp not just as a distribution channel but as a technical platform, a move we've seen other app ecosystems attempt to solidify market position. It creates a unique value proposition: developers can build AI features that are private by default, potentially avoiding regulatory and user trust pitfalls associated with data sharing.

The Performance Hurdle and Apple's Shadow

The core challenge for this partnership is performance. Liquid AI's CEO, Ramin Hasani, acknowledged the competition, noting that "Apple already provides its own local models to developers." His argument is differentiation through customization and efficiency.

"Before training our models, we select an architecture that is different and tailored to the hardware," Hasani told TechCrunch. "That allows us to really have the most efficient version of intelligence that runs directly on the device."

The key technical claim: Liquid AI's models are designed to be "adaptable and become more intelligent over time" through user input. Combined with Mnemos's persistent memory, this suggests a system that learns and improves locally, a significant step beyond static, pre-loaded models.

The XOOMAR Inference: This is a bet on architecture over raw scale. Liquid AI's models are already used by companies like Mercedes-Benz and Shopify for on-device tasks, suggesting a focus on specialized, efficient inference rather than competing with massive cloud LLMs on general knowledge. Success depends on proving these models can handle the complex, multi-step "agentic workflows" Kosovan mentioned, entirely offline.

A Credit-Based Future for AI App Stores

MacPaw isn't just rebuilding the tech stack, it's rethinking the business model. The company is "already experimenting with credit-based pricing for the app store, where users can perform a certain number of AI operations based on the credits they have and the complexity of the task."

This model, reminiscent of API pricing for cloud AI services, could apply to both local and cloud-based AI tasks within Setapp. It moves away from simple subscription fees toward metered usage, aligning cost directly with computational value. This could make sophisticated AI features more accessible within otherwise inexpensive apps, as we explored in our coverage of niche app monetization strategies.

What to watch: The developer uptake. If MacPaw can deliver a seamless SDK that simplifies the complex work of splitting tasks between local and cloud models while guaranteeing privacy, it could attract a new wave of AI-native Mac developers. The first real-world test will be the performance and user reception of Eney later this year. If it delivers a snappy, capable, and truly private experience, the platform play becomes credible. If it feels limited or sluggish, developers will likely stick with cloud APIs or wait for Apple's next move. This partnership shows that the battle for AI's future isn't just in the data centers, it's on the device in front of you.

Why This Changes Everything

• It challenges the dominant cloud-reliant AI model by offering developers a fully local, privacy-focused alternative for macOS applications.
• It addresses growing consumer privacy concerns by preventing personal data from being sent to remote servers during AI tasks.
• It could shift market dynamics in the AI assistant space by prioritizing offline functionality and native macOS integration over cloud scalability.