MacPaw has partnered with Liquid AI to integrate local, on-device AI technology into its Mac applications. This strategic move aims to improve user privacy and enable offline functionality for the company's AI assistant, Eney. MacPaw plans to eventually offer this technology to developers on its Setapp marketplace, marking a pivot toward an AI-first subscription model.
MacPaw, the developer known for its Mac applications and the Setapp subscription platform, has formed a long-term partnership with Liquid AI to bring on-device artificial intelligence to the macOS environment. The agreement centers on integrating Liquid Foundation Models with MacPaw's own inference and memory systems to create a localized AI experience.
The collaboration focuses on the development of two specific technologies: Elix, an inference engine, and Mnemos, a memory and context layer. By running AI models directly on the user's hardware rather than relying on cloud-based processing, MacPaw aims to enhance user privacy and enable its AI assistant, Eney, to operate without an active internet connection. This approach attempts to bypass the latency and privacy issues often associated with sending data to remote servers.
For MacPaw, this partnership is a significant step in transforming its business model. The company is moving toward an AI-first ecosystem, introducing a credit-based pricing system for AI operations within its Setapp marketplace. This structure allows the company to charge for tasks based on their complexity, a shift from traditional flat-fee subscription models. If successful, MacPaw plans to make this on-device inference stack available to other developers on the Setapp platform, potentially turning the marketplace into a hub for both local and cloud-based AI applications.
However, the venture faces notable business and execution risks. As both MacPaw and Liquid AI are privately held companies, they operate without the financial transparency required of public firms, relying on internal financing and venture capital to sustain long-term research and development. There is also a technical challenge in balancing model performance with hardware constraints; locally hosted models must be highly optimized to match the capabilities of large, cloud-based competitors. Additionally, the success of the new credit-based pricing model remains unproven at scale. Investors and market observers will likely monitor the performance of the Eney assistant when it launches later in 2026 and how effectively the developer community adopts these new on-device tools.