Reference

get_product_intelligence

The tool’s argument, both response shapes, and resolution semantics.

Argument

product_namestring

The product to look up. Resolution against the published catalog is typo-tolerant.

detailenum?

full (default) or compact — REST only.

Confident match

A confident match returns the full snapshot: composite scores, verified claims, pricing with live offers, per-source consensus, and 27 criteria across 7 dimensions. The resolution object reports match confidence and the other close name matches, so your agent can sanity-check what it actually got.

response (abridged)
{
  "found": true,
  "product_name": "AirPods Pro 3",
  "buyary_score": 88,
  "dimension_scores": […],
  "highlights": […],
  "considerations": […],
  "source_consensus": […],
  "offers": […],
  "resolution": { … }
}

Limited coverage

When the best match falls below the confidence floor (or has no evaluation), the tool returns found: false with did-you-mean candidates in other_matches — it suggests rather than guesses.

detail=compact

A token-lean subset: it drops the long-form prose — that’s where the tens-of-kilobyte payloads live. Scores, claims, consensus rows with stance, prices, and resolution all survive, and the shape is identical in both modes. If your agent only renders the structured subset, compact is the cheaper default.

Full field tables live in the data model.