MacPaw Enables On‑Device AI for App Store Developers with Liquid AI

MacPaw, the Ukrainian software house behind the popular Mac cleaning tool CleanMyMac, announced a strategic partnership with Liquid AI to bring on‑device inference to developers building for its App Store. The collaboration centers on a stripped‑down version of MacPaw’s AI assistant, Eney, that runs entirely offline using Liquid AI’s lightweight models. By embedding inference at the edge, developers can add conversational features, contextual recommendations, or real‑time code assistance without relying on cloud APIs, dramatically reducing latency and data‑privacy concerns.

The technical core of the integration is Liquid AI’s cross‑platform SDK, which provides a C++ inference engine wrapped for Swift, Objective‑C, and even Python. The SDK ships with pre‑quantized transformers that fit within a few hundred megabytes of RAM, allowing macOS devices to execute inference in under 150 ms for typical user queries. Developers can pull the SDK via Swift Package Manager:

let package = Package( name: "EneyEdge", dependencies: [ .package(url: "https://github.com/liquidai/edge-sdk.git", from: "1.2.0") ], targets: [ .target(name: "EneyEdge", dependencies: ["LiquidAI")] ] )

Once added, a developer can instantiate the model with a single call:

import LiquidAI let model = try! LLMModel(configuration: .eagerQuantized, modelName: "eney‑small") let response = try! model.generate(prompt: "Explain git rebase in one sentence") print(response)

Beyond the SDK, MacPaw is publishing a set of sample projects on GitHub that illustrate best practices for sandboxing the model, handling user‑data encryption, and falling back to cloud inference when the device runs out of memory. The samples also showcase how to integrate with MacPaw’s own App Store metadata API, enabling developers to surface AI‑generated app descriptions or dynamic onboarding tips directly inside their apps.

From an ecosystem perspective, this move signals a maturation of edge AI for desktop environments. Historically, on‑device inference has been dominated by mobile platforms where battery and connectivity constraints drive the need for local models. MacPaw’s initiative shows that the same constraints—privacy, latency, and bandwidth—are now relevant for desktop software, especially as users become more wary of sending personal data to the cloud.

The partnership also lowers the barrier for open‑source contributors. Liquid AI’s model zoo is MIT‑licensed, and MacPaw has pledged to accept community‑submitted model optimizations via pull requests. This collaborative pipeline could accelerate the creation of domain‑specific assistants—think code‑review bots for IDE plugins or personalized productivity coaches for task‑management apps.

In short, MacPaw’s on‑device Eney, powered by Liquid AI, equips developers with a ready‑to‑use edge inference stack, fostering a new wave of privacy‑first, low‑latency AI experiences on macOS. As more developers adopt this stack, we can expect a ripple effect across the App Store, with AI features becoming a baseline expectation rather than a premium add‑on.

Photo: Anthony Riera / Unsplash (https://unsplash.com/@frenchriera)