Active NWS alerts for a point or state, plus model-forward hazard risk — Official National Weather Service active alerts (public domain, attributed) for a lat/lon or two-letter state, combined with our model's forward 48h hazard probabilities (extreme heat, heavy rain, damaging wind). Pass lat+lon or state.

Weather-driven delay risk for a US airport in 6-hour blocks — Next-24h delay risk for a top-100 US airport, scored per 6-hour block from model wind gusts, convective potential, winter precip, and ceilings, with the drivers named and a one-line verdict.

Upcoming market resolutions in the next N days with current probabilities — Active markets resolving inside the horizon, sorted by resolution date with the current blended probability. An agent's event calendar: what settles soon and where it stands.

Probability of a specific weather event in a window, for any point or city — Our model's probability that a condition (temp above/below a threshold, precip amount, wind, any precip, named-storm landfall) occurs in a window up to 31 days out, with the method labeled. When a listed prediction market covers the same event, it is cross-referenced.

Federal contract awards from USAspending, normalized and filterable — Recent US federal contract awards (USAspending.gov) as clean records: vendor, awarding agency/sub-agency, value, NAICS, start date, and place of performance. Sorted by value. Filter by vendor, agency, NAICS, minimum value, or date.

Quarter-over-quarter 13F position changes for an institutional manager — Parses a fund's 13F-HR information tables and diffs the latest quarter against the prior one: new positions, exits, increases, decreases (with share deltas), and the change in top-10 concentration. CIKs may be zero-padded or not.

8-K corporate events classified by item number with extractive summaries — Material corporate events from SEC 8-K filings, deterministically classified by Item number (5.02 → leadership_change, 2.02 → results, 1.01 → material_agreement, …) with a two-sentence extractive summary and the filing URL. Filter by ticker, event type, or date.

Full-text search over SEC filings (EDGAR FTS), normalized results — Proxies SEC EDGAR full-text search (efts.sec.gov) with fair-access rate limiting and returns normalized hits: company, form, filing date, matched document, and archive URL. Upstream provides no text excerpt, so snippets describe the matched document.

Normalized insider transactions (Form 4) with flow aggregates for a ticker — SEC Form 4 filings parsed into clean insider transactions: buy/sell/award classification, 10b5-1 plan flags, per-filer roles, and aggregate net flow with notable-pattern flags. Use it to gauge insider conviction before events.

Historical performance of the edge signal: hit rate, calibration, by hazard — Aggregates stored edge signals whose markets have resolved: how often the model beat the market (hit rate), average absolute-error improvement (edge captured), a model calibration curve, and per-hazard breakdowns. model_label is explicit about which model produced the signals.

Model-vs-market divergence across every parseable weather prediction market — Compares our weather model's event probability with the market-implied probability for every weather-linked prediction market we can parse, ranked by divergence with direction, model confidence (from ensemble spread), and market liquidity. parse_coverage reports how much of the weather catalog is covered.

Degree-days, load signal, and renewables indexes for a US ISO region — Daily heating/cooling degree-days (°F-days), a normalized peak-load signal, and solar/wind generation indexes computed from load-weighted city forecasts for CAISO, ERCOT, PJM, MISO, SPP, NYISO, or ISONE, with extreme-risk flags.

Hourly point forecast for any coordinate, up to 240 hours — Hourly temperature, feels-like, precipitation probability/amount, wind, gusts, humidity, and cloud cover from our weather model for an exact lat/lon. Imperial by default (units=metric available). Longer horizons cost more.

Ensemble percentile bands (p10–p90) for temp, precip, or wind — Hourly p10/p25/p50/p75/p90 bands from the model ensemble for one variable at a coordinate. Metric units (°C, mm, kph). Use it to reason about uncertainty, not just point values.

One-call answer for a specific time window: rain odds, temps, comfort, verdict — Summarizes a ≤12-hour window at a coordinate into decision-ready numbers (rain probability, expected precip, temp range in °C, max gust, comfort index) plus a one-sentence plain-English verdict. Built for agents planning events, deliveries, and travel.

Probability time series for a market with key-move detection — Hourly or daily blended probability series (per-venue series included where listed) with volume, plus key_moves: buckets where the blend moved at least 5 points. Useful for charting and for change-point context before paying for realtime reads.

Summarized depth, spread, and 24h volume profile for one market — Execution-quality summary per venue: depth within 1c/5c of the touch, spread in cents, and an hourly 24h volume profile. Pass notional_usd for a slippage estimate. Derived analytics only — raw order books are never mirrored.

Markets with the largest probability moves over 1h/6h/24h windows — Active markets ranked by absolute blended-probability change over the chosen window, with volume surge ratios. The fastest way for an agent to find where news is landing.

Per-venue implied vs vig-adjusted odds with summarized depth for one market — Venue-by-venue comparison for a market: implied probability, vig-adjusted probability (overround removed via the quoted book), summarized depth near the touch, plus the cross-venue spread and a disagreement flag when it exceeds 5 points.

Cheapest endpoint: liveness plus x402 payment pipeline check — Returns ok with the server time and whether this call was paid. Use it to verify your x402 client end to end before spending on data endpoints.

Blended event probability across Kalshi and Polymarket for one market — Liquidity-weighted blend of venue implied probabilities for a canonical market slug, with per-venue detail, cross-venue divergence, and the last 1h move. Resolved markets return 410 RESOLVED with the outcome. Find slugs via /v1/probability/search.

Find prediction markets by free-text query; returns slugs for other endpoints — Full-text search over the canonical market catalog (Postgres FTS + trigram in live mode). Returns slugs, blended probabilities, and match confidence. Empty result sets return 200 with a suggestion listing valid categories.

Product recalls across FDA, USDA, and CPSC in one normalized schema — openFDA enforcement reports (food/drug/device), USDA FSIS recalls, and CPSC consumer-product recalls normalized to one shape with a unified severity class (I highest — FDA/USDA classes verbatim; CPSC derived from hazard language). Newest first.