Skip to main content
Gcore MCP Server runs locally and connects to Claude Code or Cursor IDE. For browser-based testing without installation, use MCP Inspector. Before installing:
  1. Install the uv package manager from the uv documentation.
  2. Create a Gcore API token with permissions for the operations the server will perform.
The API token is displayed once and cannot be retrieved later. Store it securely and do not commit to version control.

Step 1. Install the server

Run the server in a temporary environment without persistent installation:
uvx --from "gcore-mcp-server@git+https://github.com/G-Core/gcore-mcp-server.git" gcore-mcp-server
Pin a specific version by appending the tag:
uvx --from "gcore-mcp-server@git+https://github.com/G-Core/gcore-mcp-server.git@v0.1.1" gcore-mcp-server

Step 2. Configure the AI client

Add the server to the AI client configuration file. The configuration includes environment variables for authentication and tool filtering.
Edit ~/.claude.json:
{
  "mcpServers": {
    "gcore-mcp-server": {
      "command": "uvx",
      "args": [
        "--from",
        "gcore-mcp-server@git+https://github.com/G-Core/gcore-mcp-server.git",
        "gcore-mcp-server"
      ],
      "env": {
        "GCORE_API_KEY": "your-api-key",
        "GCORE_TOOLS": "*"
      }
    }
  }
}
Claude Code supports GCORE_TOOLS=* (all tools) due to its deferred schema loading.

Environment variables

VariableRequiredDescription
GCORE_API_KEYYesAPI token for authentication
GCORE_TOOLSNoTool filter pattern (see Filter tools)
GCORE_CLOUD_PROJECT_IDNoDefault project ID for Cloud operations
GCORE_CLOUD_REGION_IDNoDefault region ID for Cloud operations

Step 3. Verify the setup

Restart the AI client to load the new configuration. The server outputs its version and the number of loaded tools when connected. Test with a simple query:
List my CDN resources
A successful response confirms the server is configured correctly.