Test it on history.Run it live.

Langbot is the data and simulation layer for every market on Kalshi and Polymarket: order-book history with outcomes and rules, the speech, official text and data releases that move them on one clock, and a simulator with the venue's real fees.

Start free

$pip install langbot-sdk

Browse every market

Kalshi yes bid, “private credit”Goldman Sachs Q3 call · 13 Oct 2026

10:3010:33 ET

One example of context, a word that moved a market: the analyst's mention doesn't count; the company's does. Scroll to play it.

A word that moved a market.

Speech is one kind of context on the clock. Goldman Sachs' third-quarter call, 13 October 2026, on the dial-in; Kalshi's market on “private credit” trades near 10¢.

10:31:04.8. An analyst says “private credit”. Under the market's rules analysts don't count, and the bid barely moves.

10:32:18.4. The company says it. That counts, and the next recorded quote is 99¢.

Analyst10:31:04.8Doesn't count

Company10:32:18.4Counts

Each mention of “private credit” on the call, with the yes bid at the word and five seconds later
Time, ET	Said by	Under the rules	Bid at the word → 5 s later
10:31:04.8	Analyst	Doesn't count	5¢ → 9¢
10:32:18.4	Company	Counts	8¢ → 99¢
10:32:26.8	Company	Counts	99¢ → 99¢

Bids from Langbot's recorded Kalshi book, about one a second; times are word ends on the dial-in; no transcript text is shown (a restricted source).

Not every market settles on a word.

Most markets settle on a text, a number or a score, not a word. Kalshi's market on the Fed holding rates at its 28 October meeting went from 70¢ at the end of August to 30¢ in the weeks after the September meeting raised them, then to 96¢ on the eve of the decision. The statement came at 2:00 pm ET and the market settled Yes. Langbot keeps this history for every market, with what moved it on the same clock.

MarketFed maintains rate, October 2026 meeting

Kalshi yes price, daily close70¢ → 96¢

ResultYes

TickerKXFEDDECISION-26OCT-H0

Three public dates on the market's clock, with the yes price that day
Day	On the clock	Yes close
16 Sep	September meeting: rates raised 25 bp	56¢
14 Oct	CPI release, 8:30 am ET	76¢
28 Oct	Statement, 2:00 pm ET: settled Yes	97¢ last trade

Daily closes from Kalshi's public API for KXFEDDECISION-26OCT-H0, fetched 2 Nov 2026: the venue's own history, not Langbot's recording. Langbot backfills this history for every market and lines the context up beside it; live order-book recording for every open market is rolling out.

One record, four uses.

History: Order books, outcomes and the rules as written for every market on Kalshi and Polymarket, sports to weather, with each market's price history backfilled from the venue and a public page of its own. Live book recording for every open market is rolling out; recorded live events already carry books about once a second.
Context: What moved it, on the same clock: speeches, hearings and earnings calls, official text such as statements and filings, data releases and results. Where words settle a market, as on mention markets, each phrase carries who said it and whether it counts under the rules.
Research: How prices moved after an event, what a position is worth before it from the market's own history, and how likely a threshold is to be reached while it is live. One consolidated probability per real-world question across both venues, with the methodology published.
Simulate and run: Replay any market's history through your own strategy with the venue's real fees and real timing before you risk money, run it on paper live, then switch it on with your own exchange keys. Langbot never trades for you.

Replay, simulate, then run.

Replay. Open any market's history: its books, outcome and rules, with what moved it on the clock; where words count, every phrase, who said it and whether it counts.

Simulate. Put a strategy through the market's recorded book at the latencies you choose, with the venue's fees, before any money moves.

Run. Then run it on paper against the live record, and switch it on with your own exchange keys. Langbot never trades for you.

JPMorgan Chase Q3 2026 earnings callPublic webcast · recorded 13 Oct 2026

Phrases on the clockBuyback · KXEARNINGSMENTIONJPM-26OCT13-BUYB

Time, ET	Phrase	Said by	Under the rules	Yes bid → 5 s
08:46:04.4	“buyback”	Analyst	Doesn't count	32¢ → 33¢
08:46:51.0	“buyback”	Company	Counts	51¢ → 99¢

The analyst's question left the price where it was; the company's answer took it to 99¢.

The market's rule

Company representatives: Count
Analysts: Don't count
Venue: Kalshi

SimulateAny market with its recorded books

Recorded event

Goldman Sachs Q3 2026Dial-in · 13 Oct 2026
JPMorgan Chase Q3 2026Public webcast · 13 Oct 2026

Strategy and timing

Built-in strategyBuy Yes on the first counting hit

Latencies to compare250 ms2.25 s5.25 s

Fills walk the recorded book after each latency; taker fee 0.07 of C·P·(1−P).Run simulation

The book after the company's “private credit”Goldman Sachs · 10:32:18.4 ET

Your latency	Yes bid on the record
250 ms	8¢
2.25 s	99¢
5.25 s	99¢

RunPaper first, then live, from your machine

Terminal

$ pip install langbot-sdk  # the SDK and the CLI
$ langbot sim run -e <event_id> -s my_bot.py  # the recorded book, your file
$ langbot run my_bot.py --mode demo  # paper, on the live record
$ langbot run my_bot.py --mode live  # your exchange key, your machine

Test, then run

ReplayOn the record, with real fees and timing
PaperOn the live record, no money
LiveYour own exchange keys

Test first,then run.

Free with the record a minute behind live, and no card. Every plan gets the same record; paid plans get it sooner.

Start free

$pip install langbot-sdk

See pricing Read the quickstart