pg-aiguide is a Python-based Model Context Protocol server and Claude plugin designed to improve how AI coding assistants generate PostgreSQL code. The repository addresses a specific problem: AI tools frequently produce outdated, suboptimal, or incomplete PostgreSQL schemas that lack modern features, proper constraints, and performance optimizations. pg-aiguide solves this by providing AI agents with curated PostgreSQL expertise, semantic search capabilities, and best-practice skills that guide code generation toward production-ready schemas.
The project offers three primary deployment modes. First, it functions as Agent Skills installable via the npx skills command, making PostgreSQL best practices available to over 40 AI coding agents including Claude Code, Cursor, Codex, and Gemini CLI. Second, it operates as a public MCP server accessible at mcp.timerdata.com/docs that provides semantic search across PostgreSQL, TimescaleDB, and PostGIS documentation. Third, it serves as a Claude Code plugin optimized for native skill support within Claude's ecosystem. This multi-deployment approach ensures accessibility across diverse AI coding environments.
The core functionality centers on two main capabilities. The search_docs tool enables unified semantic and keyword-based search across multiple documentation sources, with PostgreSQL searches scoped by version to ensure relevance. The view_skill tool exposes curated, opinionated PostgreSQL best practices covering schema design, indexing strategies, data types, constraints, naming conventions, performance tuning, and modern PostgreSQL features. A demonstration comparing Claude Code output with and without pg-aiguide showed dramatic improvements: schemas generated with the tool included four times more constraints, fifty-five percent more indexes including partial and expression indexes, PostgreSQL 17-recommended patterns, and modern features like GENERATED ALWAYS AS IDENTITY and NULLS NOT DISTINCT.
Currently, the project supports documentation and skills for TimescaleDB and PostGIS, with pgvector planned for future inclusion. The repository welcomes community contributions for additional extensions, new best-practice skills, documentation corpora, and search quality improvements.
GitGenius activity data reveals relatively low issue and pull request volume with a median response latency of 0.7 hours across four tracked items, though mean latency extends to 384.6 hours, suggesting occasional delays on some items. The most active contributors tracked include murrayju with three events, followed by Hyunji-Youn and wachawo with one event each. The repository shares overlapping contributors with roocodeinc/roo-code, alirezarezvani/claude-skills, and everyinc/compound-engineering-plugin, indicating connections within the broader AI coding assistant ecosystem. GitGenius classifies the repository across multiple domains including PostgreSQL, Artificial Intelligence, Machine Learning, Vector Databases, Embeddings, LLMs, Data Integration, Semantic Search, Database Extensions, and Time-series Data, reflecting its position at the intersection of database expertise and AI-driven code generation.
