Install and configure CDB MCP Server

Prerequisites and server installation common to all AI clients. For client-specific configuration, see Connect Claude Code or Connect Cursor IDE.

Prerequisites

Install uv

CDB MCP Server runs via uvx, which is part of the uv package manager.

curl -LsSf https://astral.sh/uv/install.sh | sh

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

After installation, restart the terminal to update PATH.

Verify the installation:

uv --version

Create a CDB API token

Create an API token with permissions matching the products to be used. See API token permissions for required roles per product.

The API token is displayed once and cannot be retrieved later. Store it securely and do not commit it to version control.

Install the server

uvx runs the server in a temporary environment without installing it permanently. The AI client handles this automatically using the uvx command in its configuration—no manual step is needed.

To verify the server starts correctly, run it directly:

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

If the AI client shows "Failed to connect to MCP server", verify that uv is installed (uv --version) and test the uvx command directly from the terminal. A missing comma or bracket in the config file also prevents loading.

Install as a global command available system-wide:

uv tool install "gcore-mcp-server@git+https://github.com/G-Core/gcore-mcp-server.git"

If gcore-mcp-server is not found after installation, run uv tool update-shell or add the uv tool bin directory to PATH.

Use gcore-mcp-server as the command in the AI client configuration instead of the uvx command.

Environment variables

Variable	Required	Description
`GCORE_API_KEY`	Yes	API token for authentication
`GCORE_TOOLS`	No	Tool filter pattern — controls which tools are loaded
`GCORE_CLOUD_PROJECT_ID`	No	Default project ID for Cloud operations
`GCORE_CLOUD_REGION_ID`	No	Default region ID for Cloud operations

Requests that return "401 Unauthorized" indicate an invalid or expired API token. Verify the token in the CDB Technical Web Portal and check that it has permissions for the requested product.

Configure Cloud defaults

Cloud tools require project_id and region_id for most operations. Setting defaults avoids specifying them in every request.

To find project and region IDs, ask the AI client:

List my Cloud projects

List available Cloud regions

Or navigate to Cloud > Projects in the CDB Technical Web Portal and check the URL for region IDs.

Once the IDs are known, add them to the env section of the AI client configuration:

{
  "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": "cloud.*",
        "GCORE_CLOUD_PROJECT_ID": "1186668",
        "GCORE_CLOUD_REGION_ID": "76"
      }
    }
  }
}

With defaults set, requests like "Create a VM with 2 vCPUs" use the specified project and region automatically.

If Cloud requests return errors about a missing project or region, add GCORE_CLOUD_PROJECT_ID and GCORE_CLOUD_REGION_ID to the env section, or specify them inline: "List VMs in project 1186668, region Luxembourg". Requests that return "429 Too Many Requests" are rate limited. Retry after the Retry-After interval when the response includes that header.