Postgres MCP Pro — Index Tuning + Safe SQL Tools

Practical Notes

• Quant: GitHub stars and forks indicate broad usage; treat this as a production-grade starting point for DB copilots.
• Quant: start in read-only, then add write access only after you can replay the same query-analysis prompt 3 times with identical outputs.

Rollout pattern

• Point it at a staging replica first, and validate schema + EXPLAIN flows end-to-end before touching production.
• Adopt a playbook: (1) schema snapshot, (2) EXPLAIN, (3) index hypothesis, (4) verify with a before/after benchmark.
• Keep a human approval rule for any action that changes indexes or constraints.

Watchouts

Avoid giving unrestricted write access on day 1. Keep credentials scoped, start with read-only, and review any index recommendations with real workload evidence.

FAQ

Q: Does it require Docker? A: No. The README shows Docker and Python installs (pipx/uv). Use Docker when you want fewer environment surprises.

Q: How do I keep it safe? A: Prefer read-only and a least-privilege DB user. Add writes only with an explicit approval gate.

Q: What is a good first task? A: Ask the agent to list schemas/tables, then run an EXPLAIN for one slow query and propose one index hypothesis.