How do I identify trading cards in Claude or ChatGPT?
Direct answer
A chat assistant can describe a card from a photo, but it does not know Cardmarket product IDs or today's prices and often mixes up printings. Add the NeoSatoshi MCP server to Claude Code, Claude Desktop, Cursor, VS Code or Codex: the assistant then sends the photo to our card recognition API and gets back each card with set, number, rarity, Cardmarket product ID and daily price, for Pokémon, One Piece, Disney Lorcana, Yu-Gi-Oh!, Magic: The Gathering, Star Wars: Unlimited and football cards. Without a key it uses a free demo key with 10 detections per day.
Why not just ask the chat assistant?
- A vision model reads the name well, but the exact printing (set, number, 1st Edition, reverse holo, parallel) is where it guesses. A guessed printing is a wrong price.
- It has no Cardmarket product IDs and no daily prices, so it cannot tell you what a binder page is worth today.
- A recognition API matches every card against a card catalog, returns alternatives when a print is unclear and leaves the price empty instead of guessing.
Claude Code, Claude Desktop, Cursor, VS Code and Codex
These apps can use MCP servers, small programs that give the assistant new tools. Ours is the npm package neosatoshi-card-mcp. In Claude Code it is one command:
bash
claude mcp add neosatoshi-cards -e NEOSATOSHI_API_KEY=YOUR_API_KEY -- npx -y neosatoshi-card-mcpIn Claude Desktop and Cursor add this to the MCP configuration and restart the app. The MCP server page has the files for VS Code and Codex too.
JSON
{
"mcpServers": {
"neosatoshi-cards": {
"command": "npx",
"args": ["-y", "neosatoshi-card-mcp"],
"env": { "NEOSATOSHI_API_KEY": "YOUR_API_KEY" }
}
}
}Leave out NEOSATOSHI_API_KEY to start with the public demo key. You need Node.js 18 or newer.
What you can ask
- "What are the cards on this binder page worth?" The assistant reads every card on the photo (all six trading card games) and adds up the Cardmarket prices.
- "Scan my folder of Lorcana photos and give me a CSV." One detection per photo, the result is a new CSV file in that folder.
- "Card 4 is wrong, it is the foil one." The assistant searches the catalog and corrects the card for free within 24 hours.
- "Which Harry Kane Topps Chrome cards exist from 2023?" Football cards are found by player, with their parallels and print runs.
ChatGPT
ChatGPT itself cannot start a local MCP server, and there is no NeoSatoshi app in ChatGPT yet. Until there is, use Codex (OpenAI's coding assistant, which takes the same MCP server), or build a custom GPT with an Action from our OpenAPI spec (authentication: API key, Bearer) and send it links to your photos. For selling, the NeoSatoshi app does the same recognition in the browser and turns the cards into Cardmarket, eBay or Shopify listings.
What it costs
- The MCP server is free and open source (MIT).
- The demo key allows 10 detections, 10 searches and 3 lookups per day per network, without an account.
- With your own key one card scan is used per detected card; photos without a card cost nothing. Start-Up, the smallest paid plan, costs 19 per month (15 per month paid yearly, the same number in EUR, USD or CHF) and includes 1,000 card scans per month.
- Correcting a wrong card is free. A catalog search or lookup that returns a card costs half a card scan.
Frequently asked questions
Does it judge card condition?
No. It identifies which card it is. On a PSA, BGS or CGC slab it reads the grade from the label, but it never estimates a grade.
Can the assistant read all my files?
Only JPEG and PNG images, checked by their content. Set NEOSATOSHI_ALLOWED_DIRS to the folders it may use. The server only writes new result files from a folder scan and never overwrites a file.
I want to call the API from my own code instead.
Then start with the scanner app tutorial and the Detection API documentation.
Further reading
Set up the MCP server
Setup for Claude, Cursor, VS Code and Codex, all tools and example chats. Works with the free demo key.
No account required for the first 10 detections.