Today, we are excited to announce the launch of the Remote MCP Server for instances.vantage.sh. Users can now seamlessly interact with instances.vantage.sh through Large Language Model (LLM) tools, like Claude and Cursor, as well as open-source agents, like Goose, that act as MCP clients, to incorporate AWS and Azure resource pricing lookups into GenAI workflows.

Since January 2021, Vantage has been the maintainer of instances.vantage.sh (formerly ec2instances.info), the most popular open-source site for understanding and comparing EC2 instances. Earlier this year, we rewrote instances.vantage.sh to modernize the site, making it faster and easier to use. Since then, we’ve been adding features and improvements, such as a newsletter for changes and new instance types, adding the China regions, sharable links, and now, the Remote MCP Server.

Model Context Protocols (MCPs) have been increasingly popular as the emerging standard for connecting Large Language Models (LLMs) to structured external data sources. Remote MCPs allow you to interact with AI agents from your browser without needing to maintain infrastructure or install any packages.

The instances.vantage.sh Remote MCP Server is a free community tool (no Vantage account required) that works out of the box with supported LLM clients, like Claude, Cursor, Amazon Bedrock, and soon ChatGPT, to interact with instances.vantage.sh data.

You can ask questions like:

• What’s the cheapest instance with 16 vCPUs and 64 GB of RAM in us-east-1?
• Compare r7g.4xlarge and m7i.4xlarge pricing and capabilities.
• Show me all Graviton-based instances with at least 128 GB of memory.
• List all instance types in eu-west-1 with local NVMe storage.
• What’s the network bandwidth for m6i.8xlarge?
• I am looking to deploy a workload in the United States, and require a baseline of XYZ CPUs and ABC Memory. I would like to be able to scale to xyz at peak hours using spot capacity. I am open to using any regions in the US. Could you create a pricing model comparison between AWS and Azure?
• Here is an export of AWS CUR data and sample utilization data of my instances for a workload currently running on the c5 instance family. I would like to modernize this to graviton. Can you recommend to me what generation and sizes to move to, and create a pricing model comparison?

Play video

For teams using Vantage’s core product, the instances.vantage.sh MCP can be combined with the Vantage Remote MCP to seamlessly reference both public EC2 data and your own private cloud cost and usage data inside the same LLM session.

Getting started takes only a few minutes and requires no installation. Just head to https://instances-mcp.vantage.sh/ and follow the instructions.

Frequently Asked Questions

1. Do I need a Vantage account to use this?

The instances.vantage.sh MCP server is free and open to the community. No Vantage account or login is required.

2. Why does it require an email?

Your email is only used to issue an API key for authentication with the MCP server. This helps us manage fair usage and protect the service from abuse.

3. Does Vantage have access to see what prompts I ask an LLM?

No. Vantage does not receive or log the prompts you ask an LLM agent. Your queries remain private between you and the LLM client you’re using. Vantage is only able to see what API calls and responses associated with your API key.

4. How is this different from the Vantage Remote MCP?

The instances.vantage.sh MCP provides public AWS EC2 instance data (hardware specs, pricing, region availability, and features). The Vantage Remote MCP, available through the Vantage product, provides access to your private cloud cost and usage data. Together, they allow you to compare public reference data with your own spend data in a single LLM workflow.

5. What are the supported MCP clients I can use this with?

The instances.vantage.sh MCP Server is compatible with any AI agents that provide MCP client functionality. At the time of this blog post, examples of supported clients include Anthropic’s Claude (desktop application and claude.ai), Azure AI Foundry, Amazon Bedrock, Google Gemini, Cursor IDE, and Block’s local AI agent Goose. See the documentation from each provider to understand minimum version requirements required to run MCP hosts for each, as well as integrate remote MCPs.