Esta página ainda não está disponível em Português (Brasil), então o texto abaixo está em inglês. Abrir a página em inglês

MCP server

The Model Context Protocol lets an AI assistant call tools. This server gives it read-only access to the site's public price API, so it can answer "which instance is cheapest" or "what changed this week" from the vendors' list prices instead of from memory.

The prices are on-demand list prices in USD, read weekly from each vendor's public price list. They are for planning: not live prices, not contracted or discounted prices, and without tax. Every row carries its vendor source and the date it was read. The server has no key and keeps no state.

Tools

MCP tools
Tool	What it does	Example question
`find_instances`	Instances with at least the given vCPU and memory, by architecture, region group, provider and category (virtual machines or GPU). Cheapest first.	Which 8 vCPU, 32 GB instance is cheapest in the India region group?
`search_ai_models`	Language model token prices per million tokens, by tier, vendor and open weights. A price the vendor does not publish is null, never zero.	List open-weight mid-tier models with their output price.
`recent_price_changes`	Catalog SKUs added, removed or changed in price, by provider and date.	What changed on AWS since 1 October?
`compare_example_workloads`	Monthly totals per cloud for the site's example factories. A cloud with complete: false is missing a service or price and is never called the cheapest.	Compare the 1,000-device example across the four clouds.
`lookup_region`	Regions by provider, country or text, with the vendor's code, location, status and source page. Certifications are not included.	Which clouds have a region in Saudi Arabia?

Not built yet: pricing your own workload inputs, and certification statements, which come from each vendor's compliance page (see compliance).

Install

The package is not published to npm yet, so you run it from a checkout of the repository. You need Node 20 or later. In packages/mcp-server, run npm ci, then point your assistant at src/index.js using its absolute path.

Claude Code

claude mcp add cloudcostcompare -- node /absolute/path/to/packages/mcp-server/src/index.js

Claude Desktop

Add to claude_desktop_config.json

{
  "mcpServers": {
    "cloudcostcompare": { "command": "node", "args": ["/absolute/path/to/packages/mcp-server/src/index.js"] }
  }
}

Cursor

Add to .cursor/mcp.json

{
  "mcpServers": {
    "cloudcostcompare": { "command": "node", "args": ["/absolute/path/to/packages/mcp-server/src/index.js"] }
  }
}

VS Code

Add to .vscode/mcp.json

{
  "servers": {
    "cloudcostcompare": { "type": "stdio", "command": "node", "args": ["/absolute/path/to/packages/mcp-server/src/index.js"] }
  }
}

Set CLOUDCOST_API_BASE to read another copy of the API, for example a local build served from web/dist.

The API behind it

The same data is available as plain JSON at /api/v1/, described by an OpenAPI file. The assistant page uses these tools too.