Apple Harbor App Lets Macs Run AI Models Privately Without Cloud Fees

Claris, an Apple subsidiary, has developed a new app called Apple Harbor that transforms Mac computers into private AI servers capable of running open-weight models without incurring cloud fees. The app is currently in a private preview stage, and Apple has removed its unlisted entry from the App Store.

The offering is aimed at AI users who currently pay per-token fees to cloud providers. However, free tools that cluster Macs already exist, so Harbor primarily shifts AI costs from monthly subscriptions to Apple hardware purchases.

Why Apple Harbor Moves AI Costs From the Cloud to the Desk

The app, previously referred to as Fabric, appeared on X, where a user posted its description and later reported that it had been taken down.

Apple is working on a new app called Harbor

Harbor turns your Mac into a private AI inference server. Run open-weight models locally on Apple silicon—your data never leaves your hardware, and nothing reaches the cloud. Connect multiple Mac computers over Thunderbolt 5 to pool… pic.twitter.com/UYJBQeX4Nf

— Aaron (@aaronp613) November 10, 2026

Claris, which is best known for its FileMaker business database, states that a single Mac can become an AI endpoint within minutes. The company also promises that there will be no per-token billing.

Data and prompts stay strictly within your physical Mac hardware.

Claris

For larger models, Apple Harbor links two or more Thunderbolt 5 Macs and combines their unified memory. This allows models that are too large for a single machine to run across the cluster.

Promotional images show a MacBook Pro with an M5 Max chip connected to two Mac minis. Another image depicts the app downloading OpenAI’s open-weight gpt-oss-20b model from Hugging Face, a public repository of models.

This shifts spending toward hardware. Ethereum co-founder Vitalik Buterin has observed that the cost of AI inference is almost entirely tied to hardware.

For Apple shareholders, this is a hardware-focused narrative. Following Apple's earnings beat driven by record iPhone sales, Harbor provides the company with a new argument for promoting high-memory Macs.

Can a One-Click Setup Beat Free Open-Source Rivals?

Harbor is entering a competitive space. Developers already use Exo to cluster Macs and LM Studio to run models on a single Mac, both free of charge.

Apple Harbor focuses on ease of setup rather than speed, combining both tasks into one application. It automatically detects connected Macs and manages model downloads without requiring command-line input.

Claris has not disclosed a price, but the sign-up form requests a company name, suggesting an initial focus on business customers.

Running models locally could also address the chatbot lock-in concerns raised by Shark Tank investor Mark Cuban, as user data remains on local machines rather than third-party servers.

Meanwhile, open-weight models continue to improve. Bloomberg Intelligence estimates that Chinese models have reduced the US AI lead to roughly 3%.

If that trend persists, more AI users could run powerful models at home, positioning Apple Harbor to place Macs at the center of that shift. A high price or a business-only release would undermine that possibility.