pg-aiguide is an Apache-2.0 MCP server and Claude Code plugin that gives AI coding agents version-aware PostgreSQL documentation and curated best-practice skills, built for developers and teams whose assistants write outdated or incomplete Postgres code.

What it is

pg-aiguide is a knowledge layer for AI coding tools, distributed in three forms: Agent Skills installed through npx skills, a public MCP server hosted at https://mcp.tigerdata.com/docs, and a Claude Code plugin published through a Claude plugin marketplace. It is written in Python, carries the Apache-2.0 licence, and lives in the MCP and Agent Skills ecosystem alongside Claude Code, Cursor, Codex, Gemini CLI, VS Code, and more than forty other agents reachable through agentskills.io. Its content is version-aware, so the guidance an agent retrieves tracks the PostgreSQL release in play rather than a single frozen snapshot.

The concrete problem is the default behaviour of AI coding tools. Their generated Postgres is often outdated, missing constraints and indexes, unaware of modern PG features, and inconsistent with real-world practice. pg-aiguide replaces the assistant's unaided recall of PostgreSQL with semantic search across the official PostgreSQL manual plus curated, opinionated skills, so the agent retrieves documented patterns instead of guessing. The README's worked example, an e-commerce schema generated by Claude Code with and without the server enabled, reports four times as many constraints, 55 percent more indexes including partial and expression indexes, PG17-recommended patterns, use of GENERATED ALWAYS AS IDENTITY and NULLS NOT DISTINCT, and cleaner naming and documentation.

Key capabilities

• Version-aware semantic search across the official PostgreSQL manual, with TimescaleDB and PostGIS manuals also indexed.
• Curated, opinionated Postgres best-practice skills installed with npx skills add timescale/pg-aiguide --skill postgres, or interactively via npx skills add timescale/pg-aiguide.
• Public MCP server at https://mcp.tigerdata.com/docs, added to any agent through a JSON mcpServers entry named pg-aiguide.
• Claude Code plugin installation through claude plugin marketplace add timescale/pg-aiguide followed by claude plugin install pg@aiguide.
• Extension ecosystem documentation, starting with TimescaleDB, with more extensions stated as coming soon.
• One-click install badges for Cursor, VS Code, VS Code Insiders, Visual Studio, and Goose.
• Documented schema-generation improvements including more constraints, partial and expression indexes, and modern identity and null-handling syntax.

Who uses it and how

• Developers running Claude Code who want Postgres guidance delivered through Claude's native skill support rather than pasted context.
• Teams working across several agents, since the same skills and MCP endpoint serve Cursor, Codex, Gemini CLI, VS Code, and other agentskills.io clients.
• Engineers writing schema DDL, where the README's comparison shows the largest gains in constraint and index coverage.
• Teams already on TimescaleDB or PostGIS, who gain extension-specific documentation alongside core PostgreSQL manuals.
• Individual developers who want the hosted endpoint without running any infrastructure of their own.

Getting started

Install the skills with npx skills add timescale/pg-aiguide --skill postgres, or add the public MCP server by pointing an agent at https://mcp.tigerdata.com/docs. Claude Code users can instead run claude plugin marketplace add timescale/pg-aiguide and then claude plugin install pg@aiguide.

How it compares

No paid products are listed among the facts, and no directly comparable PostgreSQL knowledge server is named, so pg-aiguide stands alone in this registry. It belongs to the category of documentation and knowledge-base tooling, but its delivery mechanism is an MCP endpoint and installable agent skills rather than a browsable site or a search UI for humans.

When to use it — and when not

Use it when AI assistants are already part of a Postgres workflow and their output needs to be grounded in current, versioned documentation; the hosted MCP server means no database, storage, or mail service to operate. Skip it if the work is non-Postgres, if the team does not use AI coding agents at all, or if only core PostgreSQL matters and extension coverage does — the README states extension documentation starts with TimescaleDB and that more is coming soon. One limitation worth noting: the README documents no self-hosted deployment path for the MCP server, so the hosted endpoint is the only route described.

npx skills add timescale/pg-aiguide --skill postgres

npx skills add timescale/pg-aiguide

{
  "mcpServers": {
    "pg-aiguide": {
      "url": "https://mcp.tigerdata.com/docs"
    }
  }
}

claude plugin marketplace add timescale/pg-aiguide
claude plugin install pg@aiguide

claude plugin marketplace add timescale/pg-aiguide
claude plugin install pg@aiguide

codex mcp add --url "https://mcp.tigerdata.com/docs" pg-aiguide

{
  "mcpServers": {
    "pg-aiguide": {
      "url": "https://mcp.tigerdata.com/docs"
    }
  }
}

gemini mcp add -s user pg-aiguide "https://mcp.tigerdata.com/docs" -t http

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "pg-aiguide": {
      "type": "remote",
      "url": "https://mcp.tigerdata.com/docs"
    }
  }
}

code --add-mcp '{"name":"pg-aiguide","type":"http","url":"https://mcp.tigerdata.com/docs"}'

code-insiders --add-mcp '{"name":"pg-aiguide","type":"http","url":"https://mcp.tigerdata.com/docs"}'

{
  "mcpServers": {
    "pg-aiguide": {
      "serverUrl": "https://mcp.tigerdata.com/docs"
    }
  }
}