At NVIDIA GTC Paris back in June, we announced the Nebius MCP Server, our integration that connects Claude by Anthropic or other AI chatbots to the Nebius AI Cloud infrastructure.

As AI assistants become integral to daily workflows, why not use conversational AI to manage your cloud resources too? Whether you need to check the free capacity of your clusters, view a list of VMs or get a cost analysis by project, you can ask Claude to do it for you. Here’s how this conversational approach transforms your infrastructure management workflow.

LLM-powered infrastructure insightsLLM-powered infrastructure insights

With the Nebius MCP integration, LLMs gain deep knowledge of the Nebius AI Cloud and access to your infrastructure and resources. Our MCP server implements the MCP Tools standard, allowing models like Claude to interact with our cloud, retrieve data about the instances and perform actions.

You can use the AI chat to get any information your team would normally get from the CLI or web console. Claude understands your natural language requests, translates them to CLI commands and generates tables and charts in convenient visual formats.

Leveraging advanced LLM capabilities, you can also make complex requests, like “Show me spending trends and suggest optimizations”. Claude will map your infrastructure, summarize reports, aggregate information and offer recommendations for resource management. It’s like having a dedicated DevOps expert available 24/7 to answer questions and provide insights into your AI infrastructure.

Let’s look at how this works with real queries you might use every day.

See it in actionSee it in action

AI chat is an intuitive way to ask for information, whether you’re an AI сloud power user or just need ad-hoc answers. Here are some real-world examples of prompts and responses for listing projects, managing VMs and getting insights into NVIDIA GPUs.

List your projects in a tableList your projects in a table

Get a quick overview of your projects with a prompt like “show me all my projects as a table”. Ask for your preferred formatting, whether it’s a table, CSV or XLS. Claude gets all the necessary context from the Nebius MCP Server:

Get insights about your GPUsGet insights about your GPUs

For the prompt “show all audit events for vm mks8snodegroup-e00cwafcgcsnvgs7mx-52gnq-n57st”, Claude lists the VM details and audit events, then presents a summary to give you quick insights. Here’s an example of Claude’s observations:

Ask anythingAsk anything

Make your prompts as simple or complex as you like. Everyday scenarios might include:

• Daily infrastructure checks on free capacity and cluster status

• Get a report on spending trends over the last month with cost-saving recommendations

• Generate a chart showing current cluster utilization with suggestions for optimization

These capabilities are powered by Model Context Protocol (MCP), ensuring secure, controlled access to external resources.

Understanding MCP: The technology behind our integrationUnderstanding MCP: The technology behind our integration

Model Context Protocol is an open standard developed by Anthropic that enables LLMs to securely connect to external data sources and tools. You can think of it as an API-like technology that bridges the gap between AI capabilities and real-world systems without compromising security.

The integration has three components: an LLM client (Claude Desktop), the Nebius MCP Server and the MCP protocol that allows them to communicate.

Click to expand