ref-tools-mcp

Overview

An MCP server that gives AI coding agents efficient, token-minimized access to documentation for APIs, services, and libraries. It uses intelligent search and context-aware retrieval to find exactly the documentation context an agent needs while minimizing token usage and context rot.

Key Features

• Agentic Search — Designed to match how AI models naturally search; for simple queries it makes a couple of calls, for complex ones it refines prompts iteratively, always finding the right context with minimum tokens
• Session-Aware Filtering — Tracks search trajectory within MCP sessions so repeated similar searches never return duplicate results, allowing agents to both page through results and refine prompts simultaneously
• Selective Page Fetching — When reading documentation pages, uses the agent's search history to drop less relevant sections and return only the most relevant 5k tokens, avoiding the bloat of standard web scraping

Who is this for?

Developers using AI coding agents who need accurate, up-to-date documentation context during development. Particularly valuable for those who want their AI assistant to reference official docs without consuming excessive tokens or getting stale, irrelevant information.