MCP App Store
by KeenEthics
Spellbinder icon

Spellbinder MCP Server for Claude and ChatGPT

By Spellbinder LLC

Overview

Spellbinder is a Magic: The Gathering deck builder, card database and collection manager. Connect it and Claude can search the full card catalog by name, rules text, type, color, legality and price. It can look up printings and rulings, find the cheapest tournament-eligible printing of a card, browse sets and preconstructed decks, and research tournament events, decklists, archetypes and metagame summaries. Once you sign in, Claude can also read your decks, collection, wishlist and profile. It can analyze a deck for legality, ownership and market value, rank and summarize your collection, track set progress, and compare recent Standard tournament decks against your Paper collection and a budget. Claude can add to your account but never take away from it. It can save a new deck from a decklist and add cards to your collection or wishlist. It cannot edit or delete anything you already have. Prices are market estimates that update daily. When a card has no current price, Spellbinder says so rather than guessing.

How to connect

  1. Claude: Open the Spellbinder listing in the Claude connectors directory and choose Connect. Claude listing Spellbinder exposes 35 tools here.
  2. ChatGPT: Open the Spellbinder listing in the ChatGPT app directory and choose Connect. ChatGPT listing Spellbinder exposes 41 tools here.

Tools

archetype_get

ChatGPT
Get a public archetype’s format identity, current meta share, and bounded matchup summary. The existing app matchup service covers a rolling 60 day window across locations; every win rate includes its match count.

card_batch_get

ChatGPT
Look up up to 20 cards by oracle_id in one bounded public read. Returns one preferred visible printing per unique identity with stable card and oracle IDs, canonical links, card basics, location and legality coverage, and nullable USD/TIX prices. UUID identities are normalized to lowercase. Unresolved IDs are explicit. This does not enumerate every printing or claim complete printing coverage. Results carry image_url, image_url_small, image_url_large, back_image_url and paper-only purchase_urls; digital marketplace links stay null.

card_get

ChatGPT
Get full details for one card by oracle_id (the printing-independent card identity returned by card_search). detailed adds up to 100 recent printings with stable card_ids and per-finish prices. Check printings_complete before using that page for a global printing comparison; use card_printing_get for one selected printing. image_url is a directly renderable card image, alongside image_url_small, image_url_large and back_image_url; pass include_image: true to receive the image itself as an image content block when the user wants to see the card. The attached image is the card's canonical printing. purchase_urls contain paper marketplace links only; digital marketplace links stay null.

card_printing_batch_get

ChatGPT
Look up up to 20 exact card_ids from collection, wishlist, deck or printing results in one public catalog query. Preserves requested printing identities and first-request order after lowercase deduplication, with per-finish nullable market prices. Missing or hidden IDs are unresolved, never substituted. Visible rows with missing identity or route metadata retain null fields and a null URL when no canonical link is possible. Completeness concerns only the requested IDs; use card_printing_cheapest for a global price comparison. Results carry image_url, image_url_small, image_url_large, back_image_url and paper-only purchase_urls; digital marketplace links stay null.

card_printing_cheapest

ChatGPT
Find the cheapest currently priced printing for one card in Paper (USD) or Magic Online (TIX). Default mode searches the catalog and may include non-tournament printings. For deck shopping, supply tournament_format: format legality and printing eligibility are checked before ranking all candidates. Unknown eligibility and unpriced coverage are explicit; no ineligible fallback is used. Arena has no market price. Results carry image_url, image_url_small, image_url_large, back_image_url and paper-only purchase_urls; digital marketplace links stay null.

card_printing_get

ChatGPT
Get one exact Magic card printing by its stable card_id, returned in card_get detailed. Use this after selecting a specific set and collector number; it never substitutes a canonical or newer printing. image_url is a directly renderable card image, alongside image_url_small, image_url_large and back_image_url; pass include_image: true to receive the image itself as an image content block when the user wants to see the card. purchase_urls contain paper marketplace links only; digital marketplace links stay null.

card_printings_list

ChatGPT
List exact printings for one oracle identity, newest first, with stable card IDs and per-finish prices. Results are bounded and paged; the final card_id tie-break keeps repeated pages deterministic. Results carry image_url, image_url_small, image_url_large, back_image_url and paper-only purchase_urls; digital marketplace links stay null.

card_rulings

ChatGPT
List catalog card rulings for one accessible oracle identity, newest first, with each ruling's source and publication date. Results use bounded offset pagination and stable ordering.

collection_add_cards

ChatGPT
Add cards to the authenticated user's collection by name (additive-only: adding a card already owned with identical condition/finish/language/location/price/acquisition-date increases its quantity, capped at 99 per line — this never lowers quantities or removes anything). operation_id is required: mint a fresh UUID per intended add and keep it, so a request whose response is lost can be retried with the same key. The same key and same request return the original result without adding again. Location constrains finish: arena accepts only nonfoil, mtgo accepts nonfoil or foil; invalid pairs are rejected per line in failed. Names that do not resolve to a real card are returned in unresolved, not silently dropped — tell the user about them.

collection_check_decklist

ChatGPT
Compare a decklist against the authenticated user's collection in one location: for each card, how many copies the deck needs, how many the collection owns (any printing, any finish), and how many are missing. Defaults to Paper. The go-to tool for "can I build this deck?" and "what do I need to buy?". Mainboard, sideboard, and commander quantities are combined by oracle identity; maybeboard lines are ignored. Text format: one card per line as "4 Lightning Bolt" with optional "Sideboard:"/"Commander:" headers.

collection_rankings

ChatGPT
Return a server-ranked, bounded list of the authenticated user's active collection holdings in one market. Rank by highest owned unit price or known holding value, grouped by collection item or oracle identity. Paper uses USD, MTGO uses TIX, and Magic Arena returns ownership counts with an explicit no-market explanation. Price coverage is always reported; missing prices are unknown, never zero. A plan that does not include Collection insights receives a plan_required refusal.

collection_set_progress

ChatGPT
Get the authenticated user's set completion for one explicit location: paper, arena, or mtgo. Optionally filter to a set code and page the started sets. Unique and Playset progress are available in every location. Master is null for Arena and for sets without a finish universe.

collection_stats

ChatGPT
Summary of the authenticated user's collection: distinct items, total copies, and market value, for the live collection plus starred and trashed scopes.

collection_top_movers

ChatGPT
Get the authenticated user's largest collection price changes for a daily, weekly, or monthly window. Results are split by Paper (USD) and MTGO (TIX); Arena is omitted because it has no market. Each location returns up to 10 risers and 10 fallers, with snapshot dates and an explicit top-N completeness note. Unpriced or stale trends are omitted rather than treated as zero.

deck_analyze

ChatGPT
Analyze one accessible user deck's construction where rules are modeled, catalog legality, location warnings, mana curve, collection ownership, and listed-printing market estimate. It uses the deck's saved location. A deck without a location returns ownership and market value as unavailable rather than assuming Paper or combining locations. Main deck, sideboard, and commander count; maybeboard is excluded. construction_status is not_modeled when this tool does not have trustworthy format construction rules, while catalog legality and location checks remain independent. Market value is a current catalog estimate of the deck's listed printings, not a cheapest-printing shopping quote.

deck_create_from_text

ChatGPT
Create a NEW deck for the authenticated user from plain decklist text (additive-only: this never modifies or deletes existing decks). Text format: one card per line as "4 Lightning Bolt"; optional "Sideboard:", "Commander:", and "Maybeboard:" section headers. Lines that do not resolve to a real card are still ADDED to the deck, as entries with no printing attached, and returned in unresolved_lines — tell the user which ones need review. A new deck takes the Default Visibility preference set on the account unless visibility says otherwise, so it can be public if that is how the account is set. operation_id is required: mint a fresh UUID per intended deck and keep it, so a create whose response is lost can be retried with the same key, which replays the original deck instead of creating a second one.

deck_diff

ChatGPT
Compare two decklists card by card: what only the first runs, what only the second runs, and what both run in different numbers, for each section (commander, main, sideboard, maybeboard). Use it for "how is my list different from the deck that won", "what changed between these two versions", or before deck_edit to see what bringing one list in line with another would take. Either side can be a Spellbinder deck the user can see (source deck) or a public tournament deck (source tournament). Cards are compared as cards, not printings: the same card in a different set or finish is not a difference. shared_pct is the share of the larger list the two have in common. Read-only.

deck_edit

ChatGPT
Change the cards in one of the user's own decks: add, remove, set a quantity, move between sections, or swap a printing. Up to 50 commands per call, applied in order, all or nothing. Read the deck with deck_get first and pass its revision as expected_revision; if the deck changed since, nothing is written and the current revision is returned. mode "preview" returns the resulting changes without writing; mode "apply" writes them and needs an operation_id. Every applied edit is checkpointed first and recorded in the deck's history under this app's name, and deck_undo reverses it. A card reference is a name, an oracle_id, or a card_id from deck_get; when a name matches more than one line the call is refused with the candidates, so name one by card_id or add finish. A new card is added by name or card_id. Cannot change a deck the user does not own, a deck in Trash, categories, or the deck's name and settings (use deck_update_meta).

deck_get

ChatGPT
Get a deck by id: one of the user's own decks (from deck_list), a deck shared with them, or any public/unlisted deck. format selects the output: structured (JSON sections with printing and oracle identities), text (readable decklist), mtgo, or arena (import-ready exports). The deck location is always included; detailed structured output also includes the description, a bounded category registry, and bounded per-card category ids with assignment coverage. Detailed structured rows carry image_url and back_image_url for each resolved printing. Text, MTGO, and Arena exports merge a deck's per-printing rows by name, so a merged line above 99 copies is lowered to it and listed in quantities_clipped.

deck_list

ChatGPT
List decks the authenticated user owns, or opt into owned and shared-with-me decks. Filter by ownership, name, format, visibility, location, color identity, or starred status; sort and page with offset. Detailed results include owner display, tags, engagement, and available market estimates with nullable values.

deck_trash

ChatGPT
Move one of the user's own decks to Trash. The deck leaves their lists and its public page, and is deleted for good only after the Trash retention period; until then the owner can restore it from Trash on the site. A connected app cannot restore a deck or delete one permanently. Only do this when the user asked for this deck to be deleted or trashed. operation_id is required.

deck_undo

ChatGPT
Put one of the user's own decks back to how it was before an edit a connected app made with deck_edit. Pass restore_operation_id (the operation_id of the edit to reverse) to restore the checkpoint that edit took, or omit it to restore the newest checkpoint on the deck. The current cards are checkpointed first, so an undo can itself be undone. The 20 most recent checkpoints per deck are kept. Does not undo deck_update_meta (send the previous values back) or deck_trash (the owner restores from Trash on the site). operation_id is required and is a NEW id for this undo, not the id being reversed.

event_deck_get

ChatGPT
Get the full decklist for a successful public tournament deck returned by event_get or tournament_deck_search. Structured output includes exact printing and oracle IDs where card enrichment resolved them; unresolved entries keep their names with null identities. Text, MTGO, and Arena formats use Spellbinder’s existing export behavior. Detailed structured rows carry image_url and back_image_url for each resolved printing. Text, MTGO, and Arena exports merge a deck's per-printing rows by name, so a merged line above 99 copies is lowered to it and listed in quantities_clipped.

event_get

ChatGPT
Get one successful public tournament event by event ID or canonical slug, with bounded top placement deck links when available. Top placements are selected from a bounded candidate window and may omit later rows. Event data does not include fabricated win rates.

feedback_submit

ChatGPT
Send a bug report, feature request or other feedback to the Spellbinder team from the authenticated user's account. Offer it when the user hits a problem with Spellbinder or these tools, or asks for something Spellbinder or this connector cannot do yet; draft the message with them and call it only after they agree to send it. Never call it unasked or to report your own errors without the user's say-so. The team reads every message and may reply to the account's email address. operation_id is required: mint a fresh UUID per message and resend it only to retry a call whose result you did not receive.

metagame_changes

ChatGPT
Compare a format's public metagame over the most recent period with the period of the same length just before it: which archetypes gained or lost meta share, which entered the top archetypes and which dropped out. Use it for "what changed in Modern this week". days picks the period length (7, 14 or 30; default 7). Each archetype carries its deck count and rank in both periods, so a change is never shown without its sample; a move under 0.5 percentage points is reported as steady, and an archetype whose share is unknown in either period has status unknown and no change figure. An archetype marked new or gone was outside the top 30 archetypes in one of the two periods, which is not the same as nobody playing it. Short periods in small formats have few events: read sample before drawing conclusions.

metagame_get

ChatGPT
Summarize the public Spellbinder metagame for a supported format and bounded period. Includes deck and match sample counts, platform and event filters, methodology, and optional bounded matchup cells. Rates always include their underlying counts.

precon_get

ChatGPT
Read a public preconstructed deck using a stable deck ID, 10-character short ID, or Spellbinder deck URL. Structured output includes printing and oracle IDs where resolved; unresolved cards retain their names with null identities. Prices separate exact-printing values, oracle fallbacks, and missing data. Catalog location is explicitly unknown. Text, MTGO, and Arena formats use Spellbinder export behavior. Detailed structured rows carry image_url and back_image_url for each resolved printing. Text, MTGO, and Arena exports merge a deck's per-printing rows by name, so a merged line above 99 copies is lowered to it and listed in quantities_clipped.

profile_get

ChatGPT
Get the authenticated user's Spellbinder profile: username, display name, and preferences. Useful to confirm which account is connected.

set_get

ChatGPT
Get public metadata and exact card-slot counts for one set code. Counts distinguish catalog rows, collector-number slots, finish slots, and Paper, Magic Arena, and Magic Online availability. Returns a bounded card_search handoff instead of loading every card in the set.

tournament_deck_compare

ChatGPT
Compare up to five requested public Standard tournament decklists against the authenticated user’s Paper collection and a USD budget. Main deck and sideboard copies are aggregated by oracle identity. Missing cards use the lowest known current tournament-eligible Paper quote. This assesses only the supplied candidates, not every Standard deck. Costs are estimates and exclude shipping and tax; current catalog legality may differ from legality at the historical event.

wishlist_add_cards

ChatGPT
Add cards to the authenticated user's wishlist by name (additive-only: re-adding a wanted card at the same finish and language sums the quantities, capped at 99). operation_id is required: mint a fresh UUID per intended add and keep it, so a request whose response is lost can be retried with the same key. The same key and same request return the original result without adding again.

wishlist_stats

ChatGPT
Summary of the authenticated user's wishlist: wishlist lines, total copies, market value, price coverage, and Paper, Magic Online, and Arena breakdowns. Lines distinguish saved printing, finish, and language; they are not distinct oracle cards.

archetype_get

Claude

card_batch_get

Claude

card_get

Claude

card_printing_batch_get

Claude

card_printing_cheapest

Claude

card_printing_get

Claude

card_printings_list

Claude

card_rulings

Claude

collection_add_cards

Claude

collection_check_decklist

Claude

collection_rankings

Claude

collection_set_progress

Claude

collection_stats

Claude

collection_top_movers

Claude

deck_analyze

Claude

deck_create_from_text

Claude

deck_get

Claude

deck_list

Claude

event_deck_get

Claude

event_get

Claude

metagame_get

Claude

precon_get

Claude

profile_get

Claude

set_get

Claude

tournament_deck_compare

Claude

wishlist_add_cards

Claude

wishlist_stats

Claude

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/spellbinder"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/spellbinder)