pg-aiguide

pg-aiguide provides AI-augmented PostgreSQL expertise via a public MCP endpoint. It offers semantic search across the PostgreSQL documentation (version-aware), AI-optimized “skills” that encode best practices for schema design, indexing, constraints, and modern PostgreSQL features, and an extension ecosystem reference set starting with TimescaleDB. You can query the server to obtain version-aware guidance, recommended patterns, and concrete implementation details that you can apply directly in your SQL or migrations. The service is designed to be used by any AI coding agent or as a Claude Code plugin, enabling improved accuracy and maintainability of PostgreSQL code produced by AI assistants.

To use it, point your MCP client or integration at the pg-aiguide endpoint. The server returns knowledge-backed responses, including schema suggestions, indexing strategies, constraint recommendations, and notes on modern features like identity columns, NULLS NOT DISTINCT, partial indexes, and extension usage. You can leverage the included “skills” to enforce proven Postgres patterns across your AI-generated queries and schemas, ensuring consistency with real-world best practices across versions.

If you’re using Claude Code or another plugin-enabled environment, you can install or enable the pg-aiguide MCP server to power code-completion, schema design prompts, and documentation lookups within your editor or assistant workflows. The combination of semantic search and curated skills helps AI agents generate safer, more optimal PostgreSQL code.

Prerequisites

• Access to the public MCP server at https://mcp.tigerdata.com/docs
• A client or plugin that can consume MCP endpoints (any MCP-compatible client)

Installation steps

1. Verify network access to the MCP endpoint:
   • Ensure your environment can reach https://mcp.tigerdata.com/docs
2. Configure your MCP client to use pg-aiguide:
   • If your client expects a mcpServers.json or equivalent, add: { "mcpServers": { "pg-aiguide": { "url": "https://mcp.tigerdata.com/docs" } } }
3. Start or refresh your MCP session in your client/editor/plugin according to its instructions.
4. Validate connectivity by sending a simple prompt like: “What are best practices for indexing a user_login table in PostgreSQL 14+?”

Notes

• No local installation is required for using the public endpoint. You interact with the server over HTTP/S via your MCP client.

Tips and considerations:

• The pg-aiguide server provides version-aware PostgreSQL knowledge, so specify the target PostgreSQL version in prompts when possible to get the most relevant guidance.
• Use the semantic search capability to locate exact sections of the PostgreSQL manual or to compare different feature recommendations across versions.
• If you plan to rely on the TimescaleDB extension guidance, ensure you mention any extension version and compatibility concerns in your prompts.
• When integrating with Claude Code or other plugins, you can leverage the provided plugin installation commands in the README to connect to the MCP endpoint.
• Common issues: network restrictions blocking access to https://mcp.tigerdata.com/docs, or misconfigured mcpServers.json leading to connection errors. Double-check the server URL and JSON structure in your client configuration.