Free tool · Cratesignal
Draft a data-cited Amazon listing, fast.
Drop in a product brief — name, category, three to five features, landed cost, and target MSRP. The tool runs the 30% landed-cost gross margin gate first; if your economics clear that floor, it drafts a title, five bullets, and a description with every recommendation grounded in four citation numbers from Cratesignal's kitchen + pet cohort baseline.
Result
Your listing draft will appear here.
Submit a brief that clears the 30% landed-cost gross margin gate to see the title, five data-cited bullets, and a description — with every recommendation grounded in the citation numbers on top of the result panel.
How the data is wired
The margin filter runs server-side, before any LLM call. Margin is (targetPrice − landedCost) / targetPrice expressed as a percent. If the result falls below 30%, the route returns 422 with field-level guidance on the price input — the model is never invoked and you do not pay for a draft you cannot ship.
Review velocity and competitor count come from Cratesignal's 2026-Q3 cohort baseline — kitchen 0.70 reviews/day, pet 0.45 reviews/day; competitor counts by price bucket (0–15 USD: 120, 15–30: 85, 30–50: 40, 50+: 25). Each number is rendered back on the result panel as a citation chip, and the LLM is given the same numbers as invariants — it does not get to invent or recompute them. The plain-text disclosure on the result panel says exactly where the figures come from.
Voice rules are folded into the system prompt at src/lib/listing/prompt.ts: no emojis, no superlatives, no aspirational copy, every bullet cites at least one numeric. If you tweak the model — temperature, vendor, baseline year — update the snapshot date on the result-pane footer so the citation provenance stays honest.
For questions or to flag a baseline drift, write to cratesignal-3@polsia.app.