Nutanix rolled out a Model Context Protocol server for its Cloud Platform on August 10, giving tools like GitHub Copilot, Claude Code, and Cursor a way to actually touch cloud infrastructure instead of just talking about it in a chat window. Type a plain-language request, and the server translates it into real API calls against the Nutanix Prism V4 API, which sits at the center of how the platform manages hybrid multicloud environments.
What stands out here isn’t the AI integration itself, since plenty of cloud infrastructure vendors are racing to bolt MCP support onto their platforms right now. It’s how tightly Nutanix has boxed in what these assistants can actually do once connected to a live cloud environment.
The server runs read-only by default. Any AI tool connecting through it can only pull information, not make changes, unless an administrator explicitly flips on write access. The system logs every operation, and connected tools inherit whatever role-based permissions already exist for their Prism Central identity, rather than getting a blanket pass into the underlying cloud infrastructure. Nutanix also built in throttling and metering, plus asynchronous handling for cloud tasks that take a while to complete.
That restraint seems deliberate, and it fits the current mood around AI agents touching production cloud systems. Thomas Cornely, Nutanix’s EVP of Product Management, framed the security layer as the thing that actually makes AI-assisted cloud management usable, rather than the automation itself being the selling point.
Under the hood, the server exposes more than 1,000 API operations spanning Nutanix’s hybrid multicloud stack, and it can hand AI tools relevant schemas, code samples, and permission details. From there, developers can generate working scripts in Python, Go, Java, JavaScript, PowerShell, curl, or standard REST formats, whichever fits their existing cloud workflow.
Nutanix built detailed setup documentation for Cursor and Claude, but left GitHub Copilot’s specific configuration steps undocumented for now, even though Copilot is one of the assistants Nutanix explicitly says the server supports. GitHub’s own documentation notes that Copilot Business and Enterprise customers need to enable an MCP policy on their end before any of this works.
One caveat worth flagging clearly: this is a v0.8 technical preview, and Nutanix says outright it’s not built or supported for production cloud environments yet. Anyone testing it against a live cloud deployment should treat that warning as literal, not as standard release-notes caution.
