mcp -gravitino

This MCP server exposes a curated set of Gravitino APIs via the FastMCP integration. It provides tools for interacting with catalogs, schemas, tables, models, tags, and user-role management in a Gravitino-backed data environment. The server is started through UV in a managed virtual environment, wiring in FastMCP and HTTPX to enable efficient, asynchronous API calls. You can activate specific tools by configuring GRAVITINO_ACTIVE_TOOLS, or enable all tools with the default wildcard.

To operate, run the UV command as shown in the configuration. The server will start and expose endpoints that your clients can query to fetch metadata, assign tags, manage roles, and retrieve model information. The included tools are designed to return concise, relevant metadata suitable for downstream prompts and automations, while maintaining the semantic integrity of Gravitino APIs.

Prerequisites:

• Python 3.10+ installed
• uv (https://github.com/astral-sh/uv) installed
• Git installed

1. Clone the repository

git clone git@github.com:datastrato/mcp-server-gravitino.git

2. Navigate into the project directory

cd mcp-server-gravitino

3. Create a virtual environment using uv

uv venv

4. Activate the virtual environment

source .venv/bin/activate

5. Install dependencies

uv install

6. Run the server (example from README usage)

uv \
  --directory /path/to/mcp-gravitino \
  run \
  --with fastmcp \
  --with httpx \
  --with mcp-server-gravitino \
  python -m mcp_server_gravitino.server

Note: Adjust the directory path to your actual project location and ensure Gravitino-related environment variables are set as described in the Configuration section.

Tips and notes:

• Ensure Gravitino URI and METALAKE are correctly configured; these are required for metadata operations.
• You can switch between token-based and basic authentication by setting GRAVITINO_JWT_TOKEN or GRAVITINO_USERNAME/GRAVITINO_PASSWORD respectively.
• Use GRAVITINO_ACTIVE_TOOLS to limit enabled tools for tighter token usage; default is * (all tools).
• If you encounter connection issues, verify that GRAVITINO_URI is reachable from the host running the MCP server and that any firewall rules allow traffic.
• The MCP toolset is optimized for concise metadata responses to stay within token limits during LLM interactions.