ErrataGuard: AI Board Game Rules & Errata Resolution Agent

This is a submission for the Sanity Challenge, Path One: Ship an Agent That Queries Real Content

What I Built

Meet ErrataGuard, an AI-powered board game companion and rules agent designed to solve a problem every complex tabletop gamer faces: conflicting rulebooks, errata sheets, and developer FAQs.

Standard keyword search or generic RAG setups frequently blend rulebook text and errata updates together, resulting in hallucinated or outdated rulings. ErrataGuard uses structured Sanity schemas and Model Context Protocol (MCP) to guarantee absolute source attribution, ensuring that when an official errata overrides an original rule, both claims surface side-by-side with clear precedence ranks.

How I Used Sanity

I structured complex game components inside the Sanity Content Lake using custom schemas (ruleClause and errata), mapping out rulebook sections, version printings, and override priorities.

I then pointed Sanity Context at this structured dataset to build a navigable Knowledge Base exposed via an MCP endpoint.

• Sanity Context & MCP: The AI agent uses the Model Context Protocol (MCP) stream to query the Knowledge Base dynamically during user chat sessions.
• Structured Conflict Resolution: Rather than performing blind full-text vector searches, the agent reads structured reference documents. When an official errata updates a base rule, the agent evaluates the priority field in Sanity to surface both the original rule text and the official clarification side-by-side.

Sanity Project Details

• Sanity Project ID: xyz12345 (Production Dataset)
• Schema Model: Custom documents for ruleClause (tracking page numbers, text, version) linked via references to errata documents (tracking effective date and precedence levels).