Summary
Read-only by default, and Kafka one call away.
Queries run readonly until you deliberately set TIMEPLUS_READ_ONLY to false, which makes it reasonable to point at a live cluster. The Kafka tools are the differentiator: look at what is actually on a topic, then decide to persist it, without switching tools.
What it is
An MCP server for Timeplus. It runs SQL on the cluster, lists databases and tables, and reaches Kafka directly to list topics, sample messages and set up a streaming ETL into Timeplus.
What you get
- SQL executed against the cluster, with queries run readonly = 1 unless TIMEPLUS_READ_ONLY is set to false (run_sql)
- Databases listed, and the tables inside one of them (list_databases, list_tables)
- Kafka topics listed, and a few messages from one shown without leaving the client (list_kafka_topics, explore_kafka_topic)
- A streaming ETL that saves a Kafka topic's messages into Timeplus (create_kafka_stream)
- An Apache Iceberg database on S3 connected as a source (connect_to_apache_iceberg)
- A generate_sql prompt that gives the model Timeplus-specific SQL knowledge
Requirements
Uv, then uvx mcp-timeplus, with the Timeplus host, port, user and password in the client config. Iceberg is documented as available through Timeplus Enterprise.
Setup effort
One command plus a key — uvx mcp-timeplus, then supply credentials