For grading services
Confirm what card you're actually grading — before the cert is printed.
One REST call reads a submission photo and returns the card's name, set, number, year, and rarity in about 2-3 seconds, checked against a live catalog of 3,382,765+ cards across 43 games and sports. When it can't corroborate a match, you get an honest needsReview instead of a confident guess.
A wrong card ID on a graded cert is expensive to undo
A grader's job is to assess condition — but the intake step still has to confirm what card is actually in front of them, correctly, at volume. A misidentified parallel, a wrong set, or a swapped year printed on a cert isn't a minor typo; it's a reissue, a customer dispute, and a dent in the thing your business sells: trust in the label.
Bolting a generic image classifier onto intake doesn't solve this if it can't say 'I'm not sure.' A model that always answers with its best guess will occasionally hand you a wrong-but-confident card ID with no signal that anything was off — exactly the failure mode a grading workflow can least afford.
Built so uncertain cases surface themselves
See it identify a submission photo right now — no integration required to try it. Try the live scanner.
Priced for submission volume
Plans scale from a free tier for evaluating fit against your intake volume up to high-throughput tiers with a published uptime SLA. Rate limits rise with plan tier, and credit packs cover seasonal submission surges.
See the full pricing page for every tier, overage rate, and rate limit.
Frequently asked questions
Does this replace a human grader?
No — it identifies the card (name, set, number, year, rarity), it doesn't assess condition or assign a grade. It's built to sit at intake, before grading, so your graders start from a confirmed card identity instead of re-verifying it themselves.
What happens with a hard-to-identify parallel or variant?
You get needsReview:true and card:null rather than a forced match, with the closest candidates in alternates for a human to check. That's a normal identify request, not an error state.
Can we set our own confidence threshold for auto-pass vs. manual review?
Yes. The confidence score is calibrated to real P(correct); you decide where the auto-pass line sits for your own risk tolerance, and route everything below it — plus every needsReview result — to a person.
What's the cost at submission-center volume?
See the plan cards below for monthly caps and overage rates. For volume or SLA terms beyond the published tiers, reach out directly.
Confirm the card before you print the cert.
Get an API key, send a submission photo, and see an honest identification come back in seconds.