Spark SQL runs on your own machine — the client starts it, so there is no endpoint to ping. 108 installs a week from pypi. Last commit 9 Feb 2026.
Query Spark SQL clusters via Thrift/HiveServer2. Works with Spark, EMR, Hive, Impala.
Nothing serious here today
Today is the operative word: we check Spark SQL every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
Connect this server
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add spark-sql-mcp-server -e 'SPARK_HOST=<SPARK_HOST>' -- uvx spark-sql-mcp-server{
"mcpServers": {
"spark-sql-mcp-server": {
"command": "uvx",
"args": [
"spark-sql-mcp-server"
],
"env": {
"SPARK_HOST": "<SPARK_HOST>"
}
}
}
}[mcp_servers.spark-sql-mcp-server]
command = "uvx"
args = ["spark-sql-mcp-server"]
env = { SPARK_HOST = "<SPARK_HOST>" }{
"mcpServers": {
"spark-sql-mcp-server": {
"command": "uvx",
"args": [
"spark-sql-mcp-server"
],
"env": {
"SPARK_HOST": "<SPARK_HOST>"
}
}
}
}{
"servers": {
"spark-sql-mcp-server": {
"type": "stdio",
"command": "uvx",
"args": [
"spark-sql-mcp-server"
],
"env": {
"SPARK_HOST": "<SPARK_HOST>"
}
}
}
}Put your own value in place of
<SPARK_HOST> (Hostname of the Spark Thrift Server).
The author’s registry entry marks it as required.
Optional settings from the registry entry, left out of the block above:
SPARK_PORT (Port of the Spark Thrift Server (default: 10000)), SPARK_DATABASE (Default database to use), SPARK_AUTH (Authentication method: NONE, LDAP, KERBEROS, CUSTOM, or NOSASL), SPARK_USERNAME (Username for LDAP authentication), SPARK_PASSWORD (Password for LDAP authentication), SPARK_KERBEROS_SERVICE_NAME (Kerberos service name (default: hive)).
Alternatives to Spark SQL
same job, measured the same wayMCP server for SurrealDB: version check, SQL queries. Works with any instance.
Query an EVM contract's indexed history with SQL, via a local nuthatch nest and nuthatch dev.
MCP server for Hasura: metadata export, SQL passthrough. Works with any self-hosted instance.
Read-only health validation for Windows Server failover clusters running SQL Server FCI.
Query a TikTok data warehouse (creators, videos, sounds, daily history) with read-only SQL.
MCP server for MySQL/MariaDB — pooled async queries with destructive-query safety
Lets AI agents query SQL Server, PostgreSQL, MySQL, Oracle and SQLite via the Jam SQL Studio app.
MCP server for Trino distributed SQL query engine access
Spark SQL — questions
Answers built from our own checks of this server.