boardgame-rules-mcp

MCP server that lets AI assistants search and read board game rulebooks from 1jour-1jeu.com — a database of 12,000+ rulebooks in multiple languages.

Tools

• search_rulebook(query, language?) — fuzzy search by game name, returns up to 10 matches
• get_rulebook(game_id, language?) — downloads and returns the rulebook as text + local PDF path
• refresh_index() — re-crawls the site to update the local game index

Example Conversation

You: How many players can play Catan, and how do you win?

Claude: Let me look that up. (calls search_rulebook with query "Catan") (calls get_rulebook with game_id "catan")

Catan supports 3–4 players (5–6 with an expansion). You win by being the first player to reach 10 Victory Points. Points come from settlements (1 VP each), cities (2 VP each), the Longest Road card (2 VP), the Largest Army card (2 VP), and certain development cards.

You: What about extracting a structured summary I can reuse?

Claude: (calls get_rules_summary with game_id "catan") (receives extraction prompt, fills in the schema) (calls submit_rules_summary with structured JSON)

Done — I've cached a structured summary for Catan. Future calls to get_rules_summary will return the JSON instantly without re-reading the PDF.

Installation

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "boardgame-rules-mcp": {
      "command": "npx",
      "args": ["boardgame-rules-mcp"]
    }
  }
}

Other MCP clients

npx boardgame-rules-mcp

How it works

On first search, the server crawls the 1jour-1jeu.com sitemap and builds a local index of all games (takes a few seconds, runs once). Subsequent searches are instant. When you fetch a rulebook, the PDF is downloaded and cached locally. Extracted text is also cached so subsequent requests are fast.

Cache lives at:

• Mac/Linux: ~/.config/boardgame-rules-mcp/
• Windows: %APPDATA%\boardgame-rules-mcp\

Language codes

Use standard ISO 639-1 codes: en, fr, de, es, it, nl, pt. Not all games are available in all languages.

Troubleshooting

"PDF text extraction returned empty" The rulebook PDF is image-only (scanned pages). These cannot be extracted automatically. Try a different language if available — some editions are digitally typeset.

Game not appearing in search results The index may be out of date. Ask Claude to run refresh_index, then search again. This re-crawls the source site and takes ~30 seconds.

Wrong or outdated rulebook PDFs are cached for 90 days (configurable via BOARDGAME_CACHE_MAX_AGE_DAYS). Delete the cached file at the path shown in the get_rulebook result to force a fresh download.

Cache location

• macOS: ~/Library/Application Support/boardgame-rules-mcp/
• Linux: ~/.config/boardgame-rules-mcp/
• Windows: %APPDATA%\boardgame-rules-mcp\

License

MIT