Show HN: MCP server that teaches LLMs to write production grade Postgres SQL (github.com)

Tiger CLI (for Tiger Cloud) now bundles a Model Context Protocol (MCP) server designed to teach LLM-based assistants how to author production-grade PostgreSQL rather than acting as a thin API passthrough. The server ships with production-tested prompt templates from senior engineers, versioned Postgres docs (v15–18) and TimescaleDB material surfaced via hybrid/semantic search, and auto-discovered best practices so assistants can generate schemas and SQL with correct ID strategies, data types, case‑insensitive indexes and performance patterns. The pitch: reduce schema mistakes that trigger expensive migrations and remove manual prompt engineering—ask the assistant and get modern, production-ready Postgres code. Technically, the MCP exposes service-management and DB tools (service_list/get/create/update_password and db_execute_query with parameterized queries, timeouts and connection pooling), proxies a remote docs MCP (get_guide, semantic_search_postgres_docs, semantic_search_tiger_docs), and uses your CLI auth (tiger auth login). It supports multiple install paths (curl, brew, apt/yum, go), client integrations (Claude Code, Codex, Cursor, Gemini, VS Code, Windsurf or custom MCP clients), and can run over stdio or Streamable HTTP. Configuration lives in ~/.config/tiger/config.yaml and via TIGER_ env vars; the docs proxy is enabled by default and the project is open source under Apache 2.0. For teams building LLM-driven DB tooling, this brings contextual, safety-minded SQL generation tightly integrated with managed Postgres/Timescale services.