Skip to main content
read is the situation report. One call assembles the current picture for an asset across every domain that applies to its plane, so your agent does not stitch together price, funding, technicals, and the rest by hand. This is the call an agent makes first, and often the only one it needs.

Parameters

string
One canonical asset ID. Mutually exclusive with ids.
string list
Two to ten distinct canonical asset IDs, comma-separated. Mutually exclusive with id.

Billing

Five credits per distinct returned asset, assessment included. That is the top of the rate card and it earns it: one read carries every domain headline plus the assessment, where rebuilding the same picture from domain calls costs more credits and far more context. A batch reserves up to ten assets at the rate (50 credits) and settles the actual number delivered. An asset that is refused rather than returned is not charged. For a cheap change-watch loop, poll price at 1 credit and read on change.

A crypto read, captured live

Trimmed from the production gateway, 2026-07-16:
The assessment is deterministic characterization, not model output and not advice: positioning, structure, flow, and cascade risk, each state produced by a versioned rule (funding_oi@1) so behavior changes are visible in the data itself. risk_flags names active caveats, and confidence counts the evidence: 18 inputs proved, 0 refused. Anything that cannot be delivered moves to an unavailable block under a typed reason; a worked refusal is in the data contract.

An equity read

Equities return a headline of market and filed facts, watermarked with how fresh the filing index is. There is no assessment on the equity plane. Also live, same session:
Filed-derived values are exact decimal strings with explicit units, and the market clock (as_of) is end-of-day, separate from the filing watermark. Two clocks, both stated, neither pretending to be the other.