Predictions-as-API

The same surface VaultStore uses, behind a single bearer token.

Pre-grade forecast in JSON: a calibrated probability distribution, four sub-grades, authenticity, and EV math against your prices. We score outcomes against real grader returns and re-publish calibration daily — so when we say 87%, that means something.

Use cases

Grading services

Pre-screen submissions to reduce ungradeable return rate. Save shipping + labour on cards your model already knows are dead on arrival.

Insurance carriers

Verify declared condition before issuing a policy. Auditable distribution + authenticity score per submission.

Marketplaces

Auto-list raw inventory with a predicted grade attached. Convert browsing to bidding without forcing the seller to declare.

Sample request

Single-card forecast

POST /api/v1/oracle/predict
Authorization: Bearer vsoracle_<your-key>
Content-Type: application/json

{
  "image_urls": [
    "https://your-cdn.com/front.jpg",
    "https://your-cdn.com/back.jpg"
  ],
  "card": {
    "name": "Charizard",
    "set_name": "Base Set",
    "card_number": "4"
  },
  "current_raw_value_cents": 32000,
  "prices_by_grade": {
    "PSA 10": 380000,
    "PSA 9":  85000,
    "PSA 8":  42000
  }
}

Sample response

{
  "grade": {
    "label": "PSA 9",
    "numeric": 9,
    "distribution": { "10": 0.18, "9.5": 0.42, "9": 0.30, "8": 0.10 },
    "confidence": 0.86
  },
  "sub_grades": { "centering": 8.5, "corners": 9, "edges": 9, "surface": 9.5 },
  "authenticity": { "score": 0.97, "flags": { "reprints_suspected": false } },
  "ev": {
    "signal": "submit",
    "expected_value_after_grading_cents": 113200,
    "expected_uplift_cents": 71800,
    "psa_fee_cents": 7500,
    "shipping_roundtrip_cents": 2500,
    "slab_loss_cents": 3396
  },
  "model": { "version": "claude-haiku-…", "routed_tier": "haiku-vision" },
  "meta": { "duration_ms": 1840, "key_prefix": "vsoracle_", "key_last4": "a4f2" }
}

Pricing

Per-prediction, billed monthly

Developer

1,000 predictions / month, free during preview

Growth

100k predictions / month, calibration access, dashboard

$0.04 / call

Enterprise

Volume pricing, SLA, dedicated outcome ledger

contact

What you can’t buy from the alternatives

Calibration — every prediction is later compared against real grader returns. Pokegrade et al never see the outcome, so their confidence is a vibe. Ours converges on a number.
Distribution — full per-grade probabilities, not a single scalar. You can compute EV, value-at-risk, or any downstream financial decision yourself.
EV math — fee schedule + shipping + slab loss baked in. The response is decision-ready, not just a guess.
Population data — once a card has ≥ 25 outcomes, the API can return where this card sits in the distribution.

← Back to the Oracle

POST /api/v1/oracle/predict Authorization: Bearer vsoracle_<your-key> Content-Type: application/json { "image_urls": [ "https://your-cdn.com/front.jpg", "https://your-cdn.com/back.jpg" ], "card": { "name": "Charizard", "set_name": "Base Set", "card_number": "4" }, "current_raw_value_cents": 32000, "prices_by_grade": { "PSA 10": 380000, "PSA 9": 85000, "PSA 8": 42000 } }

{ "grade": { "label": "PSA 9", "numeric": 9, "distribution": { "10": 0.18, "9.5": 0.42, "9": 0.30, "8": 0.10 }, "confidence": 0.86 }, "sub_grades": { "centering": 8.5, "corners": 9, "edges": 9, "surface": 9.5 }, "authenticity": { "score": 0.97, "flags": { "reprints_suspected": false } }, "ev": { "signal": "submit", "expected_value_after_grading_cents": 113200, "expected_uplift_cents": 71800, "psa_fee_cents": 7500, "shipping_roundtrip_cents": 2500, "slab_loss_cents": 3396 }, "model": { "version": "claude-haiku-…", "routed_tier": "haiku-vision" }, "meta": { "duration_ms": 1840, "key_prefix": "vsoracle_", "key_last4": "a4f2" } }