PyTorch

PyTorch is an open-source machine learning framework, originally developed by Meta AI and, since 2022, stewarded by the PyTorch Foundation under the Linux Foundation. Alongside TensorFlow, it's one of the most widely used deep learning frameworks, used in both research and production, in part thanks to its dynamic computation graphs (eager execution), which make debugging more flexible than static-graph approaches.

The entries listed for PyTorch on this platform aren't MCP servers but official Claude Skills the PyTorch project publishes for its own development work — for example, writing Metal kernels for Apple hardware, triaging incoming GitHub issues, automating pull request reviews, exporting models via ExecuTorch for mobile and embedded systems, and checking type coverage with Pyrefly.

These skills are primarily aimed at contributors to the PyTorch project itself, but they're a good example of how a large open-source project can structure recurring maintenance and review work around AI agents.

Because PyTorch is governed by the PyTorch Foundation as a neutral organization under the Linux Foundation, its development isn't tied to a single company, even though Meta remains one of its largest contributors. The skills listed here reflect workflows typical of large open-source projects: without structured automation, the volume of incoming issues and pull requests would be difficult to manage.