How is this different from a futures curve alone?

Futures encode arbitrage-locked carry — what an institutional desk can't disagree with. Analyst targets encode discretionary research-desk views. Two independent forward-looking signals; the divergence between them is the signal. /v1/forward-view returns both on every call so consumers can read the disagreement directly.

Docs / Forward view

Forward view

Two-layer multi-horizon forecast for one commodity on one call. CME GC futures curve and analyst consensus from Reuters Poll, Yardeni Research, and the World Gold Council. This is the endpoint behind /data/gold.

FREE TIER+ ANALYST LAYER · PROMCP TOOL PARITY

Live endpoint

GET/v1/forward-view/{asset}Bearer · Free+

Returns every populated horizon for one commodity on one call. Asset is one of XAU, XAG, HG. Free + Basic see the futures curve. Pro adds the analyst layer (Reuters Poll + Yardeni point + WGC scenario range). Tier resolution happens at the request layer; analyst fields drop to nullfor sub-Pro callers rather than 403'ing the whole response.

Path parameters

assetstringREQUIRED: One of XAU (gold), XAG (silver), HG (copper).

Response shape

Returns a ForwardViewResponse with asset, computed_at, and a horizons[] array. Each horizon row has:

expiry — ISO YYYY-MM-DD
futures — settle price + contract code (or null if no contract matches this horizon)
analyst_consensus — Reuters Poll median (or null; Pro tier)
analyst_individual[] — Yardeni point + WGC scenario range (Pro tier)

Plus a top-level sources[] array for attribution and a scrape_metadata block for freshness.

Two layers, two questions

Each layer answers a different question; neither is ground truth. The chart at /data/gold renders both so consumers can read the disagreement directly.

Futures curve — next 5 forward COMEX gold contracts. Settle prices ~15 minutes delayed via Yahoo Finance with Stooq as a fallback. We are not a CME Group market data licensee and do not redistribute real-time CME MDP data.
Analyst consensus (Pro) — three publishers: Reuters Poll quarterly median (anchor; aggregates JPM, Goldman, BofA, Morgan Stanley, UBS, Citi, HSBC), Yardeni Research point target (independent), World Gold Council scenario range (bear / bull, NOT a confidence interval). HTML-parsed with claude- sonnet-4-5 under fair use; verbatim excerpts capped at 200 characters per row with attribution.

Refresh cadences and freshness SLAs: methodology.

MCP tool

Available as get_forward_view on every tier:

get_forward_view(asset: "XAU" | "XAG" | "HG")
  -> { asset, horizons[], sources[], scrape_metadata }

Same response shape as the REST endpoint. Tier-aware on the data layer: Free + Basic callers see analyst fields as null; Pro sees them populated when the LLM-parse pipeline has fresh data. See /docs/mcp for setup.