Mcp Postgres Analytics

What is Mcp Postgres Analytics?

Mcp Postgres Analytics is a read-only PostgreSQL analytics agent exposed as an MCP server. It connects to a production database safely—never performing any writes, DDL, or DML—and provides tools for query planning, performance diagnostics, and database statistics.

How to use Mcp Postgres Analytics?

Install via pip install git+https://github.com/archimedes-market/mcp-postgres-analytics, set the environment variable MCP_POSTGRES_DSN to a postgresql:// connection string, then run mcp-postgres-analytics serve.

Key features of Mcp Postgres Analytics

• Reads-only: no DDL, DML, or writes ever
• Startup refuses roles with write privileges (CREATE, INSERT, etc.)
• query_plan runs only EXPLAIN (ANALYZE, BUFFERS, VERBOSE)—never the query itself
• Provides tools for slow queries, index usage, table bloat, vacuum status, connections, lock waits, and size summaries
• Connects as a role with minimal pg_read_all_stats and CONNECT privileges

Use cases of Mcp Postgres Analytics

• Monitor slow queries via pg_stat_statements
• Analyze query execution plans without running the actual query
• Detect index bloat and missing-index hints
• Identify lock contention and blocking transaction chains
• Track vacuum/autovacuum progress and wraparound risk

FAQ from Mcp Postgres Analytics

Is Mcp Postgres Analytics really read-only?

Yes. The server validates at startup that the role in MCP_POSTGRES_DSN has no CREATE, INSERT, UPDATE, DELETE, TRUNCATE, ALTER, or DROP privileges. If it does, the server exits with a clear error. Additionally, the query_plan tool never executes the submitted query—it runs only EXPLAIN (...).

What database permissions are required?

The role needs only pg_read_all_stats and CONNECT privileges. No write permissions are allowed.

Which tools are available?

Seven tools: query_plan, slow_queries, index_usage, table_bloat, vacuum_status, connection_stats, lock_waits, and size_summary. Each provides production-safe diagnostic information.

How do I install Mcp Postgres Analytics?

Run pip install git+https://github.com/archimedes-market/mcp-postgres-analytics. Then set the MCP_POSTGRES_DSN environment variable to your PostgreSQL connection string and start the server with mcp-postgres-analytics serve.

Does Mcp Postgres Analytics modify the database?

No. It is designed to be a read-only agent. All tools only query system statistics and views; no data is written to the database.