What a card API business actually sells
A trading card API sells three things, and only one of them is data. The data is the card catalogue and its prices. The second is that the data stays right: new sets on release day, prices that move when the market moves, and the same card never priced as a different printing. The third is that the API answers, every time, at three in the morning, during the week a new set releases.
Customers pay for the second and third. Anyone can download a set list once. What a developer building a collection app or a shop's repricing tool cannot do on their own is keep nine games and four markets correct indefinitely. That is the business, and it is also why most attempts to build one stall.
What it takes to build from scratch
Built from nothing, a multi-game card API needs these pieces before the first customer can pay:
- A catalogue per game. Each publisher, community database and card list has its own format, its own idea of what a printing is and its own gaps. Magic alone prints one card dozens of times.
- Price feeds. Marketplaces do not publish their prices for reuse. You need a partner agreement or a scraper per market, and each one breaks when the site changes.
- Matching. The same card is listed differently on every market: per finish, per language, per reprint and per promo stamp. Matching is where most price data goes wrong, and it is invisible until a customer finds it.
- Images. A picture for every card and printing, mirrored, resized and served fast.
- The API product. Keys, rate limits, metering, billing, documentation and a dashboard.
- Operations. Monitoring, retries, blocked crawlers, broken parsers and a release every few weeks per game, for as long as the business exists.
Where these projects go wrong
Not at the start. The first game and the first market usually work within weeks. The trouble comes with the second and third: a market that blocks the crawler, a game whose numbering does not fit the schema, a reprint that gets the price of its original. Each fix is small; together they become a full-time job that pays nothing until customers arrive.
The second failure is pricing. A per-request price looks simple, but one search for a hundred cards and one lookup of a single card cost very different amounts to serve. Credit-based metering, where heavier operations cost more credits, is what the established card APIs use for that reason.
Licensing, briefly
Card names, artwork, rules text and trademarks belong to the publishers: The Pokémon Company, Wizards of the Coast, Konami, Bandai, Disney and others. Card data providers, including us, show them to identify cards, and each publisher's own policies apply to how they are used. A provider can licence you its own work, the compilation, normalisation and delivery of the data, but not the publishers' intellectual property.
If you resell someone else's API, check that your licence allows it. Most commercial licences, ours included, allow you to use data inside your own product but not to resell it as an API or a dataset without a separate agreement. That agreement is what a partner programme is.
Two routes: build it, or launch on an existing platform
Building gives you full control and costs months before the first sale, plus the permanent cost of keeping every source working. It makes sense if the data itself is your advantage: a market nobody else reads, or a kind of data nobody else collects.
Launching on an existing platform, white-label, gives you the data, the API and the operations from day one, under your own name. You trade control of the pipeline for time: with TCGGraph Partners, 5 business days from order to live. It makes sense when your advantage is somewhere else: an audience, a language, a niche, a shop network or a product people already use.
- Card data
- Yourself
A separate source for each game, each with its own format, gaps and terms. New sets arrive every few weeks across 9 games. - On TCGGraph
212,448 printings across 9 games in one schema, in 11 languages. - Prices
- Yourself
Marketplaces do not hand out their prices. Each one needs a scraper or a partner agreement, and scrapers break without warning. - On TCGGraph
Cardmarket in EUR and TCGplayer in USD, refreshed daily, plus Card Kingdom, Manapool and graded prices. - Matching
- Yourself
The same card is listed differently on every market: per finish, per language, per reprint. A wrong match prices the wrong card. - On TCGGraph
Matched printing by printing, with a status on every card that says why a price is missing. - Images
- Yourself
Pictures for 73,441 cards and their printings, to mirror, resize and serve. - On TCGGraph
Served in three widths, never metered. - The product
- Yourself
Keys, rate limits, metering, billing, docs and a dashboard, all before the first customer pays. - On TCGGraph
Included. White-label adds the portal, docs and billing in your name. - Keeping it running
- Yourself
Servers, monitoring, blocked scrapers and broken parsers, every week, for as long as the business exists. - On TCGGraph
Our job, with a 99.95% uptime commitment. - Time to first customer
- Yourself
Months, realistically, and longer for prices across markets. - On TCGGraph
5 business days.
Pricing your API
Start from what your customers do, not from your costs. A collection app syncs prices once a day; a shop reprices its stock nightly; a bot answers a few hundred lookups. Each of those is a plan with a monthly price and an allowance of credits.
Then check the margin. On TCGGraph Partners your cost is $0.70 per 1,000 credits of what your customers actually use, plus the monthly platform fee. A plan of 10,000 credits a month costs you at most $7.00 in usage, however you price it. The calculator on the partner pricing page works this out for any mix of customers and prices.
Finding the first customers
The partners who do best already know their first ten customers by name. Card shops that need prices on their website. Developers in a Discord server asking where to get data. Collectors in a country where no API speaks their language. An app with users who want an export.
Lead with one game and one kind of customer, say plainly what your API does better for them, and publish your prices. Developers decide quickly, and the ones who cannot see a price usually leave.
A 5-day launch checklist
With white-label, this is the whole list:
- Pick a name and register a domain.
- Order the white-label programme and pay the setup fee.
- Send your logo, colours and the domain you want; add the DNS records we send back.
- Connect your Stripe account and decide your plans and prices.
- Review your portal and docs, then announce it to the customers you already know.