Apple is developing a new macOS application called Harbor, discovered by leaker Aaron Perris, designed to turn Macs into private AI inference servers for running open-weight models locally on Apple silicon without cloud connectivity. Currently in private preview under the Claris brand, Harbor emphasizes zero-configuration clustering over Thunderbolt 5, automatically detecting connected Macs to pool their unified memory into a single inference resource. This architecture enables running models that exceed the capacity of any individual machine and supports higher concurrency by distributing workloads across the cluster. The app integrates model management, allowing users to browse, download, and load open-weight models directly within the interface, eliminating manual setup, command-line interaction, and infrastructure expertise. Target hardware includes Thunderbolt 5-capable systems such as Mac Studio and Mac mini, leveraging Apple silicon's unified memory architecture for efficient local inference. Apple has not announced a public release timeline, though a launch is anticipated in the coming months.