RED Atlas API

Cross-tabulates two fields (x by y) with per-cell counts and an optional metric — heatmap/matrix data (e.g. propertyType by city). EITHER AXIS MAY BE A DATE FIELD (createdAt, transactionDate, date): it is then bucketed as a calendar period via `interval`, which is how you get a per-group TIME SERIES — e.g. xField=propertyType + yField=transactionDate + interval=quarter returns quarterly counts per property type in ONE call instead of one query per period. Charges a fixed credit cost (meta.creditCost).

atlas_pivot

Aggregated statistics for a (market, entity) pair — counts, price/area aggregates, percentiles, distributions and derived metrics (e.g. `derived.pricePerBuiltAreaMedian`). Optionally add an ad-hoc `groupBy` breakdown, custom `percentiles`, or an outlier-robust `trimmed` mean. On listings and transactions the response also carries `excludingOutliers` — the same price aggregates computed without the documents ingest flagged as `isOutlier` (scraped placeholder prices such as a $99,900,000 "call for price" listing). QUOTE `excludingOutliers` FOR ANY MONEY TOTAL OR AVERAGE; the top-level `priceStats` is the unfiltered figure and can be inflated by a handful of placeholders. Charges a fixed credit cost (reported in meta.creditCost). Supports the same filters as atlas_query_data.

atlas_get_stats

THE tool for "what did comparable properties actually sell for". Returns RECORDED SALES (registry transactions — real closed prices, not asking prices) near a coordinate, each with its distance from the subject and price per m², plus the median sale price and median price per m² already computed. Optionally returns comparable active listings alongside them, clearly separated — those are ASKS and are not sale evidence. Use this for any valuation, offer-price, underwriting or "is this priced fairly" question. A near/bbox query on transactions is refused (that index has no coordinates), so this is the ONLY way to reach sales by location — never conclude comparable sales are unavailable in Atlas without calling it. PASS THE SUBJECT'S builtArea, lotArea, rooms and toilets whenever you know them: the matching is a similarity search, and comps chosen without them are merely nearby rather than comparable. Runs the comparables report and waits for it (usually seconds); if generation is still running when the wait budget ends you get status "processing" and a reportId to poll with atlas_get_report. Charges the report cost once, reported in meta.creditCost.

atlas_comparable_sales

Owner/occupant contact data (names, phones, emails, addresses) for a US/Puerto Rico address or person, via skip-trace. BEST WAY TO CALL IT: pass `parcelId` (+ market) and nothing else — Atlas then builds the query from its own records (the parcel postal address, and in Puerto Rico the owner roll's name), which matches far more often than a listing address does. Resolve the parcelId with atlas_query_data on parcels, or from a listing's parcelId field. Without a parcelId, search by ADDRESS ALONE (numbered street line + city + zip, NO name); add firstName/lastName only to narrow an address search that returned several people. The vendor matches PERSONS only — never put a company, church or organization name in lastName. Queries that cannot match (a development name with no house number, an organization, a bare surname) are REFUSED with a 400 before any vendor call, so read the error hint and fix the query instead of retrying variations. Premium: charges the contacts_enriched credit price per call (echoed in meta.creditCost) whenever contacts are DELIVERED — a no-match costs 0 credits. Repeat lookups of the same subject are served from the RED Atlas store at the same price (data.servedFrom tells which; data.resolvedQuery shows what was actually searched). Only call when the user explicitly needs owner/contact data; for parcel-linked contacts already in Atlas prefer atlas_get_parcel_contacts.

atlas_enrich_contacts

Starts a market-comparables report around a subject property: similar LISTINGS and — unless you turn it off — the RECORDED SALES near it, each with its distance from the subject. This is the Atlas source for "what did comparable properties actually sell for"; recorded sales are registry transactions, not asking prices. Async like AVM — poll atlas_get_report for the result, or call atlas_comparable_sales to do the create and the polling in one step. Charges credits per report (see meta.creditCost).

atlas_create_comparables_report

Fetches one report, including its result once status is "completed" and a shareable publicUrl. Free. Poll this after atlas_create_avm_report or atlas_create_comparables_report until status leaves "processing"; a report that ends "abstained" or "failed" was not charged and errorMessage/errorDetail say why.

atlas_get_report

Fetches one record by id for an entity type in a market. Costs credits for the fields returned; use `fields` to limit them. Includes the canonical `atlasUrl` app link for listings/parcels/properties.

atlas_get_record

Lists the entity types (properties, listings, parcels, …) available in a market and their endpoints. Free. Start here when unsure what data a market has.

atlas_list_entities

Lists the account’s generated reports (all plans). Free.

atlas_list_reports

Full market profile for ONE location (a whole market, a region, a city, a neighborhood or a sub-neighborhood): sale and lease medians with monthly series, inventory, days on market, yield, demographics, quality-of-life scores, nearby POIs and local news — in a single call. Resolve the locationId with atlas_lookup_location first. EVERY metric is an envelope: { value, unit, yoy, vsParent, series, availability, caveats } plus optional sourcePeriod { kind, start, end, label? }, captureAsOf. Prefer sourcePeriod for the value; absent means unknown. Live stock uses captureAsOf, independent of snapshot series. Supply/absorption carry inventory captureAsOf and the lagged deed sourcePeriod window. Never date surveys with meta.generatedAt. Read `availability` before quoting a number — "unavailable" means we genuinely cannot compute it (listed with the reason in meta.notAvailable), and "partial" means the value is real but the caveats change how it should be read (e.g. deed_lag_truncated_12mo, mixed_census_vintages, current_value_only). Never present a metric without its caveats, and never fill an unavailable metric from another source.

atlas_market_profile