StarRocks MCP

The StarRocks MCP server connects an AI client straight to a StarRocks cluster, covering both querying and the operational side that usually needs a separate console. read_query executes SELECT statements and other result returning commands, with an option to write output to a file for large results, while write_query runs DDL and DML that return no result set. analyze_query examines query performance using query profiles or EXPLAIN ANALYZE, which is what turns an agent from a query writer into something that can explain why a query is slow. Exploration and health tools sit on top. table_overview returns a table's structure, row count and sample rows, and db_overview does the same across every table in a database, so an agent can orient itself before writing SQL. top_hot_tables ranks the most accessed tables from audit logs and top_bad_tables surfaces the tables with the lowest health scores. query_and_plotly_chart runs a query and generates a chart from the result in one step. Alongside the tools, MCP resources expose the database list, the tables in a database, a table's schema, and internal proc paths covering frontends, backends, transactions and running jobs. The server is a Python package run through the uv package manager, configured with host, port, user and password environment variables or a single connection URL, with optional macOS Keychain storage for the password, and it supports stdio and streamable HTTP transports. It suits analytics engineers and platform teams operating StarRocks.