gcore

The Gcore MCP Server exposes a set of tools to interact with the Gcore Cloud API via language models. It uses a unified toolset configuration to selectively enable tools and patterns, allowing you to tailor the capabilities exposed to your LLM. By default, you can run the server with uvx and specify your GCORE_TOOLS environment variable to pick toolsets or patterns, and you can integrate the server into Cursor IDE (via mcp.json) for a streamlined workflow. The included example shows enabling core management and instance operations while permitting more specialized cloud or GPU-related tools through patterns.

Prerequisites:

• Node.js and npm (for MCP Inspector and optional npm tooling) or a compatible environment for uvx usage
• uv (uvx) installed and available in your PATH
• A Gcore API key with appropriate permissions (GCORE_API_KEY)

Installation and usage steps:

1. Install or upgrade uv/uvx if needed, per the official instructions: https://docs.astral.sh/uv/getting-started/installation/
2. Install or run the MCP server via uvx temporarily (one-off):

uvx --from "gcore-mcp-server@git+https://github.com/G-Core/gcore-mcp-server.git" gcore-mcp-server

3. If you want to install as a persistent tool, use:

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

4. After installation, run the command to start the MCP server (the exact command may be provided by uv after installation, e.g., gcore-mcp-server). If the command isn’t immediately in your PATH, run uv tool update-shell or ensure the tool bin directory is in your PATH.

5. To configure Cursor IDE, create or edit the configuration at ~/.cursor/mcp.json with the sample provided in the README, ensuring you replace keys and tool selections as needed.

Tips and considerations:

• The server uses GCORE_TOOLS to control which tools are exposed. You can mix predefined toolsets (like management, instances, volumes) with custom patterns (e.g., cloud.*) for granular control.
• Optional environment variables support: GCORE_BASE_URL, GCORE_CLOUD_PROJECT_ID, GCORE_CLOUD_REGION_ID, GCORE_CLIENT_ID, which can be added to your mcp.json env block if needed.
• When debugging, try the MCP Inspector (npx @modelcontextprotocol/inspector) to test tool calls and view the request/response flow.
• If you use combined tool configurations, toolsets have priority over patterns, and duplicates are removed while preserving order.
• For production deployments, securely manage your GCORE_API_KEY and avoid exposing it in logs or shared configs.