Our pick for catalog flexibility: TCGGraph
We prefer TCGGraph for a shared multi-game catalog with REST and GraphQL. Scrydex is worth evaluating when population reports, graded prices and Vision are central to your product.
The published features, side by side
| Decision point | TCGGraph | Scrydex |
|---|---|---|
| Catalog | 9 games | Multiple TCGs |
| Entry data plan | $19/month · 25,000 credits | $29/month · 5,000 credits |
| Image analysis billing | Separate recognition billing | Vision: 5 credits per request |
| Price history billing | 3 credits per history query | 3 credits per history request |
| Recognition entry option | $29/month · 6,000 matched cards | Vision included in plan features |
| Specialist emphasis | Shared catalog + REST/GraphQL | Grading + population reports + Vision |
Source-reported features and monthly USD plans reviewed during 9–13 September 2026. See source links below; recognition, tax, overage and limits can affect total cost.
Start with the output your users need
Identifying a card and analysing a collectible are related but different tasks. A scan-to-collection flow needs a usable match and a way to handle uncertainty. A grading-focused product may put more weight on graded prices and population context.
We favor TCGGraph for the first workflow because its recognition documentation describes returning a printing and prices together. Scrydex’s published package is attractive for the second. Confirm each service’s output against your required fields before committing.
Image-analysis calls are not matched cards
TCGGraph documents billing by matched card, including per-card billing in a multi-card frame. Its recognition allowance is separate from catalog credits. Scrydex lists a five-credit Vision operation within its API credit model.
To compare costs fairly, use your actual number of input images, cards per image and unresolved attempts. Do not divide the cheapest plan by its allowance and call that a universal scan price. Also budget for the catalog and history requests your workflow makes afterwards.
Evaluate recognition with your own images
Use the same sample of sleeved cards, foils, similar artwork and languages. Record exact-printing matches, unresolved results and incorrect matches separately. Set the acceptance criteria before trying either API.
This guide has not run that test and does not claim one service has better recognition accuracy. Our recommendation is about the documented product fit.
Compare a completed collection entry
Use the same labeled images and follow the result through confirmation and saving. Record the accepted printing, corrections required, returned price fields and any additional lookups. An image-analysis response and a completed inventory entry are different outcomes.
Keep difficult images in the sample: sleeves, reflective finishes, similar artwork and partial crops. Report unresolved images separately from wrong matches. A lower unresolved rate can be a poor result if it comes with more confidently incorrect printings.
Give graded data its own acceptance criteria
For slab features, test the exact grading company and grade combinations your users will add. Inspect source timestamps and missing observations. Do not let a successful raw-card scan stand in for a graded valuation test.
Use the graded pricing guide and Pokémon recognition guide to separate those requirements. Choose based on the whole product: our preference is TCGGraph for catalog flexibility, while Scrydex deserves comparison for a grading-centered workflow.
Frequently asked questions
Does this comparison prove which scanner is more accurate?
No. It compares documented scope and billing. Accuracy requires an equivalent labeled evaluation using your capture conditions.
Should Vision credits and TCGGraph scans be combined into one unit?
No. Count the images, accepted cards and downstream requests, then apply each provider’s documented rules.
Sources & review notes
Based on provider documentation reviewed during 9–13 September 2026. No independent benchmarks were performed. “Not documented” is not proof that a feature is unavailable.