v0.1.2

• Add MCP Registry metadata for discovery
• Update server.json to use PyPI package type

v0.1.1

• Publish to PyPI with pip install spark-sql-mcp-server or uvx spark-sql-mcp-server
• Add spark-sql-mcp-server CLI entry point for uvx compatibility
• Update all README config examples to use uvx
• Fix changelog workflow (create PR instead of pushing to protected main)
• Add PyPI trusted publishing workflow

v0.1.0 — Initial Release

Spark SQL MCP Server v0.1.0

An MCP server that enables AI assistants to query Spark SQL clusters via the Thrift/HiveServer2 protocol. Works with Apache Spark, AWS EMR, Hive, Impala, and Presto.

Features

• 4 MCP Tools — list_databases, list_tables, describe_table, execute_query
• Multiple Auth Methods — NONE, LDAP, KERBEROS, CUSTOM, NOSASL
• Read-Only Enforcement — Only SELECT, SHOW, DESCRIBE, EXPLAIN, and WITH statements allowed
• Safety Defaults — Automatic LIMIT on unbounded queries, sanitized error messages
• EMR Compatible — Works with AWS EMR clusters out of the box

Security

• SQL injection prevention via identifier validation
• Error sanitization — internal details (hostnames, paths, stack traces) are never exposed to clients
• Credential masking in logs and repr output
• Auth mode validation

Local Development

Docker Compose setup included for running a local Spark Thrift Server with sample data:

cd docker && docker compose up -d
pytest -m integration -v

Installation

pip install git+https://github.com/aidancorrell/spark-sql-mcp-server.git

See README for Claude Code / Claude Desktop configuration.