MacPaw and Liquid AI Plan On-Device AI Platform for Eney and Setapp Developers
MacPaw partners with Liquid AI to bring private, offline AI inference to Eney and Setapp, while planning developer tools and credit-based pricing plans.
Summary
MacPaw, the Ukraine-based software company behind the Setapp subscription store, has partnered with Liquid AI to develop locally hosted artificial intelligence for its products. Announced in a TechCrunch report published on August 5, 2026, the collaboration initially centers on Eney, MacPaw’s AI assistant unveiled in 2025. The companies are creating an on-device inference system called Elix alongside a local memory system, with the aim of processing more AI tasks directly on users’ hardware rather than relying entirely on remote servers.
Liquid AI co-founder and CEO Ramin Hasani said the company chooses and trains model architectures for particular hardware environments, seeking greater efficiency when models run on a device. MacPaw CEO Oleksandr Kosovan identified privacy, security and offline operation as intended benefits. If the system performs as described, Eney could run assistant features and agent-style workflows without a continuous internet connection. Liquid AI also plans a customization layer through which models could use user input to adapt and improve over time, although the article does not specify how that information will be stored, governed or protected.
The partnership is intended to extend beyond MacPaw’s own assistant. After the local-processing architecture has been finalized, MacPaw plans to offer the technology to developers building software for Setapp, which has more than 150,000 paying users. Kosovan said the platform would combine Liquid AI’s on-device capabilities with access to cloud-based models from providers such as Google. MacPaw’s stated goal is to give developers a single route to local and cloud inference, potentially allowing each application to select the approach best suited to a task.
The move places Setapp more directly in the emerging market for AI software distribution. Apple already supplies local models to developers, so MacPaw and Liquid AI will need to demonstrate meaningful advantages in performance, customization or developer convenience. From an industry perspective, the hybrid design could matter because local processing can reduce cloud dependence and support offline use, while cloud models can handle jobs that exceed a device’s capabilities. Setapp’s existing paid audience gives MacPaw a base through which to test that proposition, but the report provides no adoption forecasts or comparative performance data.
MacPaw is also experimenting with credit-based Setapp plans under which AI operations would consume credits according to the task and its complexity. That could help account for the uneven computing costs of different AI features, but neither credit allowances nor prices have been disclosed. The next milestones are completion of the Elix architecture, integration with Eney and eventual developer access. Important uncertainties include the launch schedule, supported hardware, model capabilities, privacy controls, developer fees and whether credit-based billing will supplement or replace parts of Setapp’s current subscription structure.
Positives
- The Elix system is being designed to run AI inference on users’ devices, enabling Eney assistants and agentic workflows to operate offline.
- MacPaw plans to pair on-device models with a local memory system, which could reduce dependence on remote processing for personalized AI functions.
- Setapp’s base of more than 150,000 paying users gives developers an established commercial audience for future AI applications.
- MacPaw intends to give Setapp developers access to both Liquid AI’s local models and cloud services from providers such as Google through one platform.
- Liquid AI says it selects architectures for specific hardware, with efficiency, privacy and security presented as core design priorities.
Risks & concerns
- MacPaw has not disclosed a release date for Elix, the locally hosted version of Eney or general developer access.
- The article provides no benchmark results to substantiate Liquid AI’s performance claims against Apple’s existing local models or other alternatives.
- MacPaw has not revealed the price, credit allowance or task-conversion rules for Setapp’s experimental credit-based AI plans.
- Liquid AI’s proposed model customization will use user input, but the report does not explain consent, retention, security or data-governance controls.
- Apple already offers local models to developers, creating competitive pressure for MacPaw and Liquid AI to prove that their stack offers a clear advantage.