Setting Up the Compare Cloud Costs MCP Server: The Complete Guide

Hey everyone, Rodrigo here!

As AI coding assistants like Claude Desktop, Cursor IDE, Visual Studio Code (Copilot, Cline, Roo Code), and Windsurf become central to daily engineering workflows, developers frequently ask their agents questions like:

"What is the cheapest 8 vCPU, 32 GB RAM virtual machine in Frankfurt across AWS, Azure, GCP, and DigitalOcean?"
"Can you compare Anthropic Claude 3.7 Sonnet, OpenAI GPT-4o, and DeepSeek V3 in terms of prompt caching and output token pricing?"
"Estimate the monthly multi-cloud infrastructure cost for a 3-tier web application or a production RAG pipeline."

Until recently, AI assistants either guessed, hallucinated outdated rates from obsolete training data, or forced developers to leave their editor to consult bloated cloud calculators.

To fix this, we officially released the Compare Cloud Costs MCP Server.

Using the open Model Context Protocol (MCP) standard, our server connects your local AI assistants directly to Compare Cloud Costs' production pricing catalog—providing real-time access to over 27,900+ active pricing SKUs across 24 cloud and AI providers, fully normalized to the FinOps FOCUS™ v1.0 standard.

Here is the complete step-by-step guide to installing, configuring, and leveraging the MCP server in your development environment.

   ┌─────────────────────────────────────────────────────────────┐
   │ AI Assistant (Claude Desktop / Cursor / VS Code / Windsurf) │
   └──────────────────────────────┬──────────────────────────────┘
                                  │ JSON-RPC (stdio transport)
                                  ▼
   ┌─────────────────────────────────────────────────────────────┐
   │  @comparecloudcosts/mcp-server (Local Protocol Engine)      │
   └──────────────────────────────┬──────────────────────────────┘
                                  │ HTTPS (X-API-Key)
                                  ▼
   ┌─────────────────────────────────────────────────────────────┐
   │  Compare Cloud Costs Production API (27,900+ Live SKUs)     │
   └─────────────────────────────────────────────────────────────┘

📦 Where the MCP Server is Published

Our MCP server is open-source, publicly registered, and distributed across primary package managers and directories:

Channel	Link / Identifier	Verification Status
npm Registry	`@comparecloudcosts/mcp-server`	Active (v1.0.0, stdio transport)
GitHub	`github.com/rodrigo-orzari/ccc-mcp-server`	Official repository with `server.json` manifest
mcpservers.org	`mcpservers.org/servers/comparecloudcosts-com-mcp`	Approved and indexed
Glama Directory	`glama.ai/mcp/servers/rodrigo_orzari/ccc-mcp-server`	Verified listing via in-repo `glama.json`
Web Portal	comparecloudcosts.com/mcp	1-Click Key Generator & Live Schema Explorer

🛠️ The 4 Available Tools

When your AI assistant connects to the server, it automatically discovers 4 specialized tools:

1. `compare_cloud_pricing`

Queries and compares compute, database, storage, and serverless pricing across AWS, Azure, Google Cloud, Oracle Cloud Infrastructure (OCI), DigitalOcean, Alibaba Cloud, and Cloudflare.

Parameters: category (required: vm, database, serverless, storage, ai, gpu, containers, app-hosting), provider, min_vcpu, min_ram_gb, region.
Sorting: Results are automatically returned ordered by price ascending (cheapest first).

2. `search_ai_models`

Searches token pricing (input, output, and prompt caching rates) and context window limits across frontier AI model providers: OpenAI, Anthropic, DeepSeek, Mistral, Meta Llama, and Cohere.

Parameters: provider, model_family.

3. `estimate_workload_architecture`

Generates end-to-end multi-cloud monthly cost benchmarks for realistic production architectures (compute + managed DB + load balancers + object storage + egress).

Parameters: workload_id (e.g. 3-tier-web, rag-ai-app, serverless-web-app, k8s-app-platform, streaming-analytics, ecommerce-microservices).

4. `lookup_datacenter_compliance`

Audits regional data center presence, sovereign cloud availability, and geographic locations across 24 global infrastructure providers.

Parameters: compliance_standard (e.g. SOC 2, ISO 27001, HIPAA, FedRAMP, DORA), country_or_city.

🚀 Step 1: Generate Your Free API Key

Tool executions require an API key to communicate with the Compare Cloud Costs catalog:

Visit comparecloudcosts.com/mcp.
Click Generate preview key (instant, 1-click generation, no credit card required).
Copy your live API key, which will look like:
```
ccc_live_xxxxxxxxxxxxxxxxxxxxxxxx
```

(Note: tools/list discovery works without an API key; executing tools/call requires CCC_API_KEY set in the environment).

⚙️ Step 2: Client Configuration Walkthroughs

The server runs locally via npx (requires Node.js ≥ 18). Add the server configuration to your preferred tool below:

1. Claude Desktop

Open your Claude Desktop configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Linux: ~/.config/Claude/claude_desktop_config.json

Add the compare-cloud-costs entry under mcpServers:

{
  "mcpServers": {
    "compare-cloud-costs": {
      "command": "npx",
      "args": ["-y", "@comparecloudcosts/mcp-server"],
      "env": {
        "CCC_API_KEY": "ccc_live_YOUR_KEY_HERE"
      }
    }
  }
}

Save the file and restart Claude Desktop.

2. Cursor IDE

You can configure Cursor either through the GUI or directly via mcp.json:

Via GUI:

Open Cursor and navigate to Cursor Settings → Features → MCP.
Click + Add New MCP Server.
Set Name to compare-cloud-costs.
Set Type to command.
Set Command to:
```
npx -y @comparecloudcosts/mcp-server
```

Via Project or User mcp.json: Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json (workspace root):

{
  "mcpServers": {
    "compare-cloud-costs": {
      "command": "npx",
      "args": ["-y", "@comparecloudcosts/mcp-server"],
      "env": {
        "CCC_API_KEY": "ccc_live_YOUR_KEY_HERE"
      }
    }
  }
}

3. Visual Studio Code (Roo Code, Cline, Copilot)

Add the server definition to .vscode/mcp.json in your project root:

{
  "servers": {
    "compare-cloud-costs": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@comparecloudcosts/mcp-server"],
      "env": {
        "CCC_API_KEY": "ccc_live_YOUR_KEY_HERE"
      }
    }
  }
}

4. Windsurf / Codeium Cascade

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "compare-cloud-costs": {
      "command": "npx",
      "args": ["-y", "@comparecloudcosts/mcp-server"],
      "env": {
        "CCC_API_KEY": "ccc_live_YOUR_KEY_HERE"
      }
    }
  }
}

💬 Real-World Prompts to Try

Once configured, your assistant will seamlessly invoke Compare Cloud Costs tools during conversation. Here are a few prompts you can try:

"What is the cheapest 8 vCPU, 32 GB RAM virtual machine in Europe across AWS, Azure, GCP, and DigitalOcean?"

Your assistant will call compare_cloud_pricing with category: "vm", min_vcpu: 8, min_ram_gb: 32, region: "europe" and format a side-by-side comparison.

"Compare Claude 3.7 Sonnet, GPT-4o, and DeepSeek V3 in terms of prompt caching and output token pricing per million tokens."

Your assistant will call search_ai_models to fetch verified token pricing and context limits.

"What would our monthly infrastructure cost be if we deployed our application using the 'rag-ai-app' architecture model on AWS versus GCP?"

Your assistant will call estimate_workload_architecture with workload_id: "rag-ai-app" to return itemized line items across compute, vector databases, and egress.

🔒 Security & Enterprise Best Practices

Zero IAM Credentials: The MCP server never touches your AWS, Azure, or GCP accounts. It communicates strictly with public rate cards, meaning you never expose cloud credentials or production read access to your assistant.
Stdio Transport: The server executes as a child process of your IDE over standard input/output (stdio), ensuring your environment variables and session tokens remain local to your machine.
Catalog Freshness: Compare Cloud Costs refreshes its pricing catalog approximately once a week via automated ingestion pipelines. It is designed for architecture modeling, budgeting, and FinOps comparison rather than sub-second spot market trading.

🤝 Questions & Contributing

The MCP server is fully open-source under the MIT license:

Source code, issue tracker, and contribution guidelines are hosted on GitHub (rodrigo-orzari/ccc-mcp-server).
Have a feature request or need support for a custom enterprise setup? Check out our Documentation or reach out directly on the About Page.

Happy building and happy optimizing!
Rodrigo Orzari
Founder, Compare Cloud Costs

Setting Up the Compare Cloud Costs MCP Server: The Complete Guide

Connect Claude Desktop, Cursor, VS Code, and Windsurf to 27,900+ cloud pricing records with Model Context Protocol.

Setting Up the Compare Cloud Costs MCP Server: The Complete Guide

📦 Where the MCP Server is Published

🛠️ The 4 Available Tools

1. compare_cloud_pricing

2. search_ai_models

3. estimate_workload_architecture

4. lookup_datacenter_compliance

🚀 Step 1: Generate Your Free API Key

⚙️ Step 2: Client Configuration Walkthroughs

1. Claude Desktop

2. Cursor IDE

3. Visual Studio Code (Roo Code, Cline, Copilot)

4. Windsurf / Codeium Cascade

💬 Real-World Prompts to Try

🔒 Security & Enterprise Best Practices

🤝 Questions & Contributing

1. `compare_cloud_pricing`

2. `search_ai_models`

3. `estimate_workload_architecture`

4. `lookup_datacenter_compliance`