Kalshi Weather Markets Scraper: Temperature, Rain, Hurricanes & Tornadoes, 28 Cities + Weather Stations

$2.00 per 1,000 rows. No API key. No proxies. Runs in under two minutes.

Pulls every open weather market on Kalshi (daily high and low temperature, hourly temperature, heat streaks, rain, snow, hurricanes, tropical storms, tornadoes, earthquakes and climate markets) and returns them as flat, spreadsheet-ready rows with the city, weather station, storm name, bracket bounds and settled outcome already parsed into their own columns.

It also reads settled markets, including the observed temperature that decided them, and can walk Kalshi's deep archive back to August 2021.

---

What you get

A single flat row per market. No nested JSON to unpick, no cent-vs-dollar guessing, no reading rules text to find out which city a ticker belongs to.

ticker	city	station	date	bracket	yes_bid	yes_ask	implied_prob	volume
KXHIGHNY-26AUG29-T79	New York City	CLINYC	2026-08-29	<79	0.56	0.57	0.565	2242.64
KXHIGHNY-26AUG29-B79.5	New York City	CLINYC	2026-08-29	79-80	0.36	0.37	0.365	715.79
KXRAIN-26AUG28-CHI	Chicago	CLIORD	2026-08-28	>0 inches	0.00	0.01	0.005	4188.00

Storm markets, which have no city at all, come back keyed on the storm instead:

ticker	storm_name	storm_metric	category	wind threshold	basin	yes_ask
KXHURCAT-26DOLLY-T1	Dolly	hurricane_category	1	74 mph	Atlantic	0.09
KXHURCAT-26DOLLY-T2	Dolly	hurricane_category	2	96 mph	Atlantic	0.02

And settled markets carry what actually happened:

ticker	date	bracket	result	observed	settlement
KXHIGHNY-26AUG27-T87	2026-08-27	>87	no	77 °F	$0.00

---

30-second quickstart

1. Click Start. The defaults fetch three representative series, about 226 rows, in a few seconds: a temperature ladder, a multi-city rain event and a storm bloc.
2. To widen it, clear Series tickers entirely for every weather series with open markets, about 1,280 rows in roughly six minutes. To narrow it instead, list the tickers you want. Series are discovered at runtime, so new ones appear without an update.
3. To get results instead of prices, set What to fetch to settled.

---

Coverage

Measured on a live full run:

• 109 series with open markets, out of 360 in Kalshi's catalogue; the other 251 are renamed duplicates and unlaunched tickers, filtered out automatically.
• 28 cities, 41 distinct weather stations, 57 named storms, 17 regions and ocean basins.
• 99.7% of rows resolve a city, a named storm or a region. The remainder are markets whose rules genuinely state no location.

Every city Kalshi lists is in the United States. These are the 28, and there are no others. The scraper reads whatever the exchange publishes, so if a city is missing it is missing from Kalshi, not filtered out here:

Atlanta, Austin, Boston, Chicago, Dallas, Dallas/Fort Worth, Denver, Hoover Dam, Houston, Houston Hobby Airport, Las Vegas, Los Angeles, Louisville, Miami, Minneapolis, New Orleans, New York City, Newark, Oklahoma City, Philadelphia, Phoenix, San Antonio, San Diego, San Francisco, San Jose, Seattle, Trenton, Washington DC.

Kalshi writes several of these two ways: Austin and Austin, TX, Central Park, New York City and New York City. Both spellings are preserved in city and unified in city_canonical, and the cities filter matches either, so you never have to know which wording a given market used.

Outside the United States

There is international coverage, but it is not city-level. Those markets are about a basin, a country or the whole planet, so they carry a region instead of a city. Filter them with regions, not cities:

region	What it covers
Atlantic, Eastern Pacific, Central Pacific	Hurricanes, named storms, season totals
Worldwide	Global temperature records, CO2 concentration, earthquakes, volcanoes
Japan	Earthquake markets
European Union, India	Climate-policy markets
Arctic	Sea ice extent
Pacific	ENSO / El Niño (RONI)

// Atlantic hurricane season markets
{ "mode": "live", "regions": ["Atlantic"] }

Market types, by open-market count on a recent run: storms 347, low temperature 291, high temperature 288, rain 156, hourly temperature 63, heat streaks 50, global temperature 17, earthquakes 17, ENSO 16, sea ice 9, CO2 5, volcanoes 4, climate policy 4.

---

Input

Every field is optional.

Field	Default	What it does
mode	live	live for open markets and prices, settled for outcomes, both.
seriesFilter	KXHIGHNY, KXRAIN, KXHURCAT	Which series to fetch. Clear it entirely to fetch every weather series that currently has open markets.
includeStorms	true	Include hurricane, tropical storm and tornado markets.
includeOrderbook	false	Add resting bid depth, a derived yes-ask and the spread. Batched, so it costs very few extra requests.
includeRecentTrades	false	Add recent executed trades. Costs one request per market.
historyDays	7	Settlement lookback window.
includeHistoricalArchive	false	Also walk Kalshi's deep archive, back to 2021.
incremental	false	Remember the newest settlement seen and fetch only newer ones next run.
maxResults	5000	Safety cap on rows per run.
requestsPerSecond	2	Politeness limit.

Filters

Ask for the slice you want instead of downloading everything and grepping it. Filters run before rows are written, so a narrower request returns fewer rows and costs less.

Field	What it does
cities	Keep only these cities: ["Chicago", "New York City"]. Case- and punctuation-insensitive, and matches both the city as Kalshi wrote it and its canonical name, so archive rows written "Central Park, New York City" still match "New York City".
regions	Keep only these regions, basins or countries: ["Atlantic"]. This is how you filter storm markets, which have no city.
stations	Keep only these weather station codes: ["CLINYC", "CLIORD"].
stormNames	Keep only these named storms: ["Yolanda"].
metrics	Keep only certain market types (high_temp, rain, storm, …).
minStrike / maxStrike	Strike window in the market's own unit: °F for temperature, inches for rain, category for hurricanes. A market is kept when any part of its range overlaps the window, so open-ended markets like "below 79" are not silently dropped.
dateFrom / dateTo	Weather-date window, YYYY-MM-DD. This is the date the market is about, not its close date.
minVolume	Drop markets that have traded less than this. The most effective single lever for cutting a run down to markets anyone is actually trading.
minOpenInterest	Drop markets with fewer open contracts than this.
minProbability / maxProbability	Implied-probability window, 0–1. 0.4–0.6 finds near coin-flips; maxProbability: 0.05 finds longshots.
results	Settled modes only. Keep only yes, no or scalar outcomes.

// Every Chicago rain market anyone is actually trading
{ "mode": "live", "cities": ["Chicago"], "metrics": ["rain"], "minVolume": 100 }

// Near coin-flips across every weather market
{ "mode": "live", "minProbability": 0.4, "maxProbability": 0.6 }

// New York highs between 70 °F and 80 °F this week
{ "mode": "live", "cities": ["New York City"], "minStrike": 70, "maxStrike": 80 }

---

Scheduling recipe

Kalshi's public data is cached for 15 seconds, so polling faster than that returns identical bytes. A sensible schedule:

• Prices: hourly, mode: live, with a seriesFilter covering only the cities you trade. A three-series run is about 10 requests and 60 rows.
• Results: once daily after settlement, mode: settled with incremental: true. The first run fetches your lookback window; every run after that fetches only what is new, and finishes in seconds.

Pipe the dataset straight into Google Sheets, S3 or a webhook using Apify's built-in integrations, no extra code.

---

What makes this different

Storm markets are covered. Hurricanes, tropical storms and tornadoes are the single largest bloc of open weather markets on Kalshi. They carry a storm name and an ocean basin rather than a city, so tools built around "city + temperature" return empty columns for all of them. This one parses the storm name, basin, category, wind threshold and season window.

Multi-city rain markets are decomposed. Kalshi packs 22 cities into a single rain event, and the city appears only on the individual market. Rows come back with the right city and station on each.

The observed temperature is recovered. Over half of Kalshi's archived settled markets have an empty observed-value field. Because a daily temperature ladder has exactly one winning bracket, the real reading can be bounded from it. The actor fills observed_value_min and observed_value_max, and marks whether the value was reported or derived. That lifts observed-value coverage on a five-year New York pull from 45% to 90%.

Five years of history joins cleanly. Kalshi's rules wording changed over time ("Central Park, New York" in 2022, "New York City (CLINYC)" in 2026), and its tickers were renamed (HIGHNY → KXHIGHNY, KXLOW → KXLOWT). city keeps the published text; city_canonical gives one name you can group by across the whole archive.

Nothing is hardcoded. Series, cities and stations are all discovered at runtime. Kalshi has renamed tickers twice and moved its settlement source once; a scraper with a baked-in city list breaks on the next change.

Strike direction is read, not guessed. A -T79 ticker is a less-than market and a -T86 is a greater-than one. The letter means nothing. Bounds come from the payload, with explicit floor_inclusive / cap_inclusive flags.

---

Output fields

Identity: ticker, event_ticker, series_ticker, market_title, event_title, series_title, category, schema_version

Location: city, city_canonical, station, station_kind, city_source, region, region_kind

Storms: storm_name, storm_basin, storm_metric, storm_category, storm_wind_threshold_mph, season_start, season_end

Classification: weather_metric, unit, metric_source, frequency, tags, market_date, mutually_exclusive

Strike: strike_type, floor_strike, cap_strike, floor_inclusive, cap_inclusive, bracket_label, yes_sub_title, no_sub_title

Prices: yes_bid, yes_ask, no_bid, no_ask, last_price, previous_price, previous_yes_bid, previous_yes_ask, yes_bid_size, yes_ask_size, implied_probability_yes

Size: volume, volume_24h, open_interest, liquidity, notional_value

Settlement: status, result, settlement_value, settlement_ts, observed_value, observed_value_min, observed_value_max, observed_value_source

Times: open_time, close_time, expiration_time, expected_expiration_time, can_close_early, created_time, updated_time, scraped_at

Optional: orderbook_yes_bids, orderbook_no_bids, orderbook_best_yes_bid, orderbook_best_no_bid, orderbook_yes_ask, orderbook_yes_ask_size, orderbook_spread, recent_trades

Raw: rules_primary, rules_secondary are passed through untouched, so you can always check the parsing against the source.

---

Notes on accuracy

• Prices are dollars per contract, 0.0000 to 1.0000. A missing quote is null, never 0.
• Volumes and sizes can be fractional. They are numbers, not strings.
• result can be yes, no or scalar. scalar means a voided or partial settlement. Do not assume every settled market pays 0 or 1.
• Settled markets can later be disputed or amended. updated_time is included so you can upsert rather than append.
• city_source and observed_value_source tell you where each derived value came from, so you can filter to only what was read directly from the payload.
• Order books are bid-only on both sides. The orderbook_yes_ask field is derived as 1 − best_no_bid, and it matches Kalshi's own quote on every market checked.

---

Schema stability & versioning

Every row carries schema_version. Fields will be added, never silently renamed or removed within a major version. Kalshi replaced its integer-cent price fields with decimal-string dollar fields in 2026 and broke most tools in this niche; this one reads the current format and fails loudly rather than emitting zeros.

---

Compliance

This is an unofficial tool. It is not affiliated with, endorsed by, or connected to Kalshi in any way. "Kalshi" is used only to describe what the data is about.

It reads Kalshi's public, unauthenticated market-data endpoints at a deliberately conservative 2 requests per second with backoff and a circuit breaker. It does not log in, place orders, or access any account data.

You are responsible for your own use of the data, including compliance with Kalshi's terms. Nothing here is trading or investment advice.

---

Changelog

1.0: First release. Live and settled modes, deep archive, storm and region parsing, multi-city rain decomposition, observed-value recovery, order book depth, incremental runs.

---

Keywords

kalshi weather api, kalshi weather markets, kalshi temperature markets, KXHIGH data, KXHIGHNY, KXLOWT, KXLOW, KXRAIN, KXHURCAT, kalshi hurricane markets, kalshi tornado markets, kalshi rain markets, weather prediction market data, prediction market scraper, kalshi scraper, kalshi api, temperature futures data, weather trading data, kalshi settlement history, kalshi historical data, event contracts weather