alibabacloud-hologres

Hologres MCP Server provides a universal interface between AI Agents and Hologres databases, enabling automated metadata retrieval and SQL execution. It exposes a suite of built-in tools that can list schemas and tables, fetch table DDLs, run SELECT and DML statements, and even generate execution and query plans. The server can be run locally via the UV/UVX bridge and is configured through MCP client files to connect to a Hologres instance using host, port, user, password, and database credentials. Once running, AI Agents can call these tools to perform data discovery, validation, and data manipulation tasks against Hologres without embedding database logic directly in the agent.

Key capabilities include executing SELECT queries (execute_hg_select_sql), performing DDL and DML operations (execute_hg_ddl_sql, execute_hg_dml_sql), retrieving table statistics (gather_hg_table_statistics), and obtaining query/execution plans (get_hg_query_plan, get_hg_execution_plan). For metadata and schema discovery, there are tools to list schemas and tables (list_hg_schemas, list_hg_tables_in_a_schema) and to show DDLs for tables, views, or external tables (show_hg_table_ddl). Built-in resources and templates enable convenient access to schemas, tables, and their DDL/statistics via hologres:/// and hologres:// URLs, as well as system paths for various diagnostics.

To use, configure the MCP client with the provided hologres-mcp-server entry, supply the required HOLOGRES_* environment variables, and invoke the available tools through your agent’s tooling layer. The server handles the authentication and translation to Hologres SQL under the hood, returning results suitable for decision-making and data-driven prompts.

Prerequisites:

• Python 3.8+ (for pip mode) or UV/uv (for UV-based mode) installed on your system
• Access to a Hologres database (host, port, user, password, database)
• Internet access for installing packages if using pip or UV/uvx

Mode 1: Using Local File (UV)

1. Clone the repository locally (or download the package).
2. Ensure you have UV installed and accessible in your PATH.
3. Run the MCP integration config (as shown in the README) with the appropriate directory path to the local repository.

Mode 2: Using PIP Mode

1. Install the MCP server package via pip:

pip install hologres-mcp-server

2. Configure MCP client with one of the supported modes (uv or uvx) as shown in the README, using HOLOGRES_* environment variables to connect to your Hologres instance.

Additional installation tips:

• If you are behind a firewall, ensure outbound access for Python package installation and any required database connectivity.
• Use a dedicated user in Hologres with appropriate permissions for the operations you intend to perform (SELECT, DDL, DML, etc.).
• Store credentials securely; avoid embedding plain credentials in MCP config files for production.

Environment variables HOLOGRES_HOST, HOLOGRES_PORT, HOLOGRES_USER, HOLOGRES_PASSWORD, and HOLOGRES_DATABASE must be set to connect to your Hologres instance. When using uvx for installation, provide the hologres-mcp-server executable name directly in the command arguments. If you encounter connectivity issues, verify network access to the Hologres host/port and confirm that the user has necessary permissions. The command and environment variable placeholders in the MCP config should be replaced with real values in production deployments. Review the list of tools and templates to determine which ones you will expose to your AI agents and ensure proper authorization controls for potentially dangerous actions like DDL or DML.