1. Normalize
Product title, seller, variant, price, stock, review and price-history signals are normalized by Scoutly AI.
Claude API integration
This page documents the planned Claude API workflow for turning structured product signals into transparent buy, wait or watch decision support.
Workflow
The Scoutly product layer collects and normalizes product data. Claude receives structured, non-sensitive commerce signals and returns explanation-ready output with evidence, uncertainty and next trigger logic.
Product title, seller, variant, price, stock, review and price-history signals are normalized by Scoutly AI.
Claude classifies the decision as buy, wait or watch and explains the tradeoff in plain language.
Generated notes are checked against source signals so the output does not overstate certainty.
Users receive a decision-support note and a next trigger, not a guaranteed purchasing outcome.
API contract
The production endpoint will keep API keys server-side and call Claude from a backend route. The front-end prototype currently displays the same response shape with demo data.
{
"product": "Sony WH-1000XM5",
"current_price": 298,
"thirty_day_low": 289,
"stock_state": "limited",
"seller_count": 3,
"review_signal": "stable",
"price_history": "near recent low, scarcity rising"
}
{
"decision": "BUY",
"confidence": "medium-high",
"evidence": [
"Price is within 3% of the recent low",
"Stock pressure is rising",
"Seller count is still sufficient"
],
"risk_label": "availability risk",
"note": "This is a strong buy window, but not a guaranteed best price.",
"next_trigger": "Watch for price rising above $319 or seller count dropping below 2"
}