Cloud Cost MCP

What jasonwilbur/cloud-cost-mcp MCP server does

The jasonwilbur/cloud-cost-mcp MCP server gives an MCP client a common interface for comparing infrastructure prices across AWS, Azure, Google Cloud, and Oracle Cloud Infrastructure. Its comparison functions cover virtual machines, object and block storage, data transfer, and managed Kubernetes. It can also identify the lowest-cost provider for requested compute specifications.

For broader estimates, the server combines compute, storage, egress, and Kubernetes inputs into a monthly workload calculation. Preset-based estimates cover scenarios such as web applications, APIs, databases, data lakes, Kubernetes clusters, high-traffic sites, and GPU workloads. Migration analysis compares an existing provider with a target provider and estimates potential savings.

How it works

The jasonwilbur/cloud-cost-mcp MCP server uses bundled pricing data for regular comparisons and can refresh provider data through public APIs. AWS, Azure, and GCP data comes from instances.vantage.sh and provider-specific sources; Azure also supports the Azure Retail Prices API, while OCI data comes from Oracle’s public pricing API. Refresh results are cached in memory for 60 minutes, according to the project documentation.

Tools report pricing data by provider, region, instance family, category, or storage tier where applicable. AWS tools include EC2, RDS, Lightsail, regions, and instance families. GCP and Azure tools provide region and instance-family or category listings. OCI tools include service categories and GPU shape data. Status and freshness tools help determine whether APIs are reachable and whether bundled data may be stale.

Setup and configuration

Install the npm package with:

npx -y cloud-cost-mcp

For Claude Desktop, add a server entry with npx as the command and -y, cloud-cost-mcp as its arguments, then restart the client. The README also documents a global npm installation using npm install -g cloud-cost-mcp, followed by the installed cloud-cost-mcp binary in the MCP client configuration.

No cloud-provider API keys or other credentials are required. The refresh functions call publicly available pricing endpoints, so access to those endpoints is required when using live refresh operations.

Tools and capabilities

The jasonwilbur/cloud-cost-mcp MCP server provides tools for:

• Comparing VM prices by vCPU and memory across four providers.
• Comparing storage tiers and data-transfer costs.
• Comparing EKS, AKS, GKE, and OKE costs.
• Estimating complete workloads and preset deployments.
• Calculating possible migration savings.
• Refreshing AWS EC2 and RDS, Azure, GCP, and OCI pricing data.
• Listing cloud regions, instance families, Azure categories, and OCI categories.
• Reviewing OCI GPU shapes, specifications, prices, and price-per-GPU comparisons.
• Checking API availability and bundled-data freshness.

The README lists more than 2,700 instance types in the current comparison description and describes broader coverage across provider data sources. Counts can vary as source data changes.

Limitations and notes

Pricing is for estimation and informational use. Cloud prices can change and may differ by region, account, operating system, licensing model, commitment term, usage pattern, and other provider-specific conditions. The project specifically identifies OCI BYOL variants in OCI refresh results and warns about stale data after 30 days.

Results should be checked directly against the relevant provider before making purchasing or architectural decisions. Availability also depends on the public pricing APIs being accessible when a live refresh is requested.