gcp-cost

The GCP Cost MCP Server exposes a set of tools to help you estimate Google Cloud costs directly from AI assistants. Available tools include: get_estimation_guide, which dynamically generates an estimation guide for any GCP service by analyzing SKUs; list_services, which lists all supported Google Cloud services with their IDs; list_skus, which returns SKUs (billable items) for a specific service; get_sku_price, which provides pricing details for a specific SKU; and estimate_cost, which computes a cost estimate for a given SKU and usage amount with automatic free tier deduction. These tools are designed to be used in a chain or in parallel to explore pricing, assemble requirements, and then obtain an actionable cost estimate. You can invoke them from supported MCP clients (Claude Desktop, Cursor, Gemini CLI) by selecting the gcp-cost MCP server and following the prompts to specify the service, region, and usage as needed. The workflow typically starts with get_estimation_guide to gather required parameters and factors, then uses estimate_cost (and optionally list_services/list_skus/get_sku_price) to refine inputs and deliver a final estimate.

Prerequisites:

• A supported MCP client (Claude Desktop, Cursor, Gemini CLI, etc.) configured to use an MCP server named gcp-cost.
• Internet access enabled for fetching public pricing data.

Installation options:

Option A: Homebrew (macOS/Linux) — Recommended

1. Ensure Homebrew is installed.
2. Install the MCP server:

brew tap nozomi-koborinai/tap
brew install gcp-cost-mcp-server

The binary will be installed to an appropriate path (e.g., /opt/homebrew/bin/gcp-cost-mcp-server or /usr/local/bin/gcp-cost-mcp-server).

Option B: Download pre-built binary

1. Download the appropriate binary from GitHub Releases for your platform.
2. Make it executable:

chmod +x gcp-cost-mcp-server-<platform>

3. (Optional) rename to gcp-cost-mcp-server and move into your PATH.

Option C: Build from source

1. Install Go 1.21+.
2. Clone and build:

git clone https://github.com/nozomi-koborinai/gcp-cost-mcp-server.git
cd gcp-cost-mcp-server
go build -o gcp-cost-mcp-server .

4. Place the binary somewhere in your PATH.

Configure the MCP client after installation:

Claude Desktop example:

{
  "mcpServers": {
    "gcp-cost": {
      "command": "/path/to/gcp-cost-mcp-server"
    }
  }
}

Cursor example:

{
  "mcpServers": {
    "gcp-cost": {
      "command": "/path/to/gcp-cost-mcp-server"
    }
  }
}

Gemini CLI example:

{
  "mcpServers": {
    "gcp-cost": {
      "command": "/path/to/gcp-cost-mcp-server"
    }
  }
}

Note: Replace /path/to/gcp-cost-mcp-server with the actual binary path on your system.

Tips and common issues:

• The server uses public pricing data; no Google Cloud project setup is required for the pricing data itself.
• The get_estimation_guide tool automatically pulls SKUs and pricing factors to generate dynamic guides, including free tier deductions where applicable.
• When using estimate_cost, you can explicitly provide a SKU_id and usage_amount. The tool will apply the free tier where supported.
• If you encounter path or permission errors, ensure the gcp-cost-mcp-server binary is executable and reachable from your MCP client configuration.
• Environment variables are not strictly required, but you can set a GCP_API_KEY or equivalent only if your deployment environment requires authenticated access to pricing catalogs (typical setups rely on public data and do not need credentials).
• You can upgrade via Homebrew (brew upgrade gcp-cost-mcp-server) or download a newer binary release as needed.
• If using multiple services in parallel, you can run several get_estimation_guide and estimate_cost calls concurrently to speed up multi-service comparisons.