MTG Deck Manager MCP Server

MTG Deck Manager MCP Servers

The MTG Deck Manager MCP servers bridge the gap between Claude and the intricate world of Magic: The Gathering. They expose a rich set of deck‑management operations—uploading full decks, drawing cards, managing hands, performing mulligans and sideboarding—as well as direct access to the Scryfall API for card searches, random draws, and detailed card data. For developers building AI‑powered tools or conversational assistants around MTG, this server eliminates the need to write custom parsers or maintain separate APIs; Claude can now treat deck operations as first‑class actions, just like any other external tool.

At its core, the Deck Manager component handles all lifecycle events of a deck. Users can upload a text list, after which the server parses and stores it in memory. Subsequent requests let Claude draw a specified number of cards, automatically updating the remaining deck size and moving drawn cards into an active hand. Mulligan logic is supported, allowing a player to discard the current hand and redraw from a reduced deck. Sideboarding—adding or removing cards before a match starts—is also available, giving developers the ability to simulate tournament preparation workflows. Because every operation is exposed through MCP, Claude can orchestrate these steps in a conversational context: “I want to mulligan my hand and then draw five more cards,” and the assistant will execute the entire sequence seamlessly.

The Scryfall API integration brings powerful card lookup capabilities to the same conversational surface. Users can request a random card, search by name or set, and retrieve complete JSON objects describing rarity, mana cost, rules text, and artwork URLs. This integration is invaluable for deck building assistants that need to suggest card substitutions or verify legality. By exposing Scryfall endpoints via MCP, developers can compose complex queries—such as “find all red creatures with power 3 or less” –and receive structured responses that Claude can immediately present to the user.

Typical use cases include building a virtual play‑testing environment where an AI coach guides a player through deck construction, simulating draws and suggesting optimal mulligans. Another scenario is integrating the server into a chatbot that assists tournament players with sideboard decisions or deck sanity checks. Because MCP servers are stateless and can run alongside Claude’s own runtime, developers can scale the deck manager independently, adding caching or persistence layers without changing the assistant’s prompts.

Unique advantages of this implementation lie in its lightweight design and tight coupling with the Scryfall ecosystem. The server is written in Python using fastmcp and httpx, ensuring low latency responses. It also supports both synchronous deck operations and asynchronous API calls, allowing Claude to interleave user instructions with background card data retrieval. For developers familiar with MCP, the server’s clear separation of deck logic and external API access makes it easy to extend or replace components—such as swapping Scryfall for another card database—without rewriting the core conversational flow.