Build on every prediction market

One normalized API and SDK. Change the venue parameter to reach a different venue.

GET/api/{exchange}/fetchMarkets

Swap that one segment for any of the 16 product venues:

• 
• 
• 
• 
• 
• 
• 
• 
• 
• 
• 
• 
• 
• 
• 
• 

Quickstart

API reference

Pick your path

Three ways in, depending on what you are building.

Read market data

Catalog, prices, and history across venues.

1. Quickstart — key, first request, pagination
2. Venue coverage — what each venue actually supports
3. Historical data — OHLCV candles and current depth

Build a trading bot

Read with a typed client, then sign and submit your own orders.

1. Quickstart — key, first request, error envelope
2. TypeScript SDK — the same surface, typed
3. Trading & execution — opt-in, client-side signing, venue status

Build an AI agent

The docs are machine-readable; the API is agent-ready.

1. AI agents hub — choose the chat-client or API route
2. llms.txt — the whole corpus indexed for retrieval
3. Agent skill — a published, versioned skill

One contract, complete infrastructure

Unified market access

One normalized method family across venues: fetchMarkets, fetchMarket, fetchEvents, fetchSeries, fetchOrderBook, and fetchTrades. Swap the {exchange} path segment — the response shape stays normalized, while unsupported venue/verb combinations fail honestly. Markets and events carry asOf + provenance; markets add capabilities. Order books carry asOf + provenance + sourceMetadata.

Historical data

Cross-venue OHLCV via fetchOHLCV (12 resolutions from 1s through 1d, up to 5,000 candles per call). Every candle is labeled with its source and quality — you always know whether you are looking at a true venue candle or a point-derived one. Coverage is availability-dependent; no uninterrupted capture or per-venue freshness guarantee is made. See Historical data for the current per-venue depth.

Cross-venue intelligence

Clusters of equivalent markets matched across venues, plus indicative price discrepancies, typed market relationships, and executable analysis only when every live-price, depth, fee, status, and resolution gate passes. See Cross-venue intelligence.

Trader Intelligence

Venue-scoped trader activity, profiles, leaderboards, versioned scores, and a smart-money feed built from public venue-published or on-chain evidence. Unsupported venue concepts stay explicitly unsupported. See Trader Intelligence.

Execution

Isolated hosted execution and client-side signing for supported venues. The normalized read API never receives a builder’s private key, and every venue publishes its exact capability and custody model. See Trading.

Simple credit pricing

Requests are endpoint-weighted — catalog reads cost 1 credit, live order-book reads 5, history 5. Sign-up is free: every account starts with 25,000 monthly credits. Trading ships on every plan; arbitrage, price-gap, and bulk endpoints start at Builder. See Credits & billing.

Every prediction market, one integration

Universal market access is the destination. Today, the normalized data contract covers 16 served product venues. The table retains the implemented Smarkets contract as a marked dark venue. The {exchange} column is the path segment — swap it to reach a different served venue.

Venue	{exchange}	Venue	{exchange}
Polymarket	polymarket	Rain	rain
Polymarket US	polymarket_us	PredictFun	predictfun
Kalshi	kalshi	SX Bet	sxbet
Smarkets (dark)1	smarkets	Pascal	pascal
Opinion	opinion	XO Market	xo
Hyperliquid	hyperliquid	PRED	pred
Limitless	limitless	PredictStreet	predictstreet
Myriad	myriad	Novig	novig
Gemini	gemini

A router pseudo-exchange unions the list verbs across all served venues.

Execution state differs per venue and changes independently of this page — GET /v1/exec/venues is the live truth. As documented: Pascal, PredictStreet, and XO Market are armed isolated execution lanes; XO advertises build + submit + cancel, PredictStreet has build + submit with no cancel, and PRED is darked, its whole lane unregistered, so even build answers a 404. All four serve catalog and order books (plus a trades tape on Pascal and PredictStreet), none has a hosted account lane, and Pascal and PredictStreet have proven venue-history coverage.

The venue coverage matrix tells you exactly what each venue supports: real CLOB depth vs. reconstructed books, and which venues expose a public trades tape.

Built for developers and AI agents

Four clients over one contract, plus a REST surface every one of them speaks.

TypeScript SDK

Typed client over the normalized contract, with signing kept client-side. TypeScript guide

Python SDK

The same surface for research and backtesting workflows. Python guide

MCP server

Thirty-three read, intelligence, and platform tools plus ten guardrailed execution tools, all registered by default. MCP guide

WebSocket API

Live capability-qualified order-book, trade, and reference-feed streams. Streaming guide

REST API — exchange-style verbs, one error envelope, cursor pagination. The API reference documents the contract every client above is built on.

Enterprise SQL — read-only analytical access over the normalized catalog and relationship model. Requires a dedicated sql scope that no self-serve plan grants; available on request.

Universal interface, capability-honest

One normalized contract does not mean every upstream venue exposes every verb. Books from venues without a real order book are flagged synthetic; missing public trades or trader identity answer an honest NOT_SUPPORTED; execution is exposed only for documented venues; and cross-venue price gaps remain indicative unless every executable gate passes.

Footnotes

1. Dark venue. Smarkets is implemented but not served on this deployment: every /api/smarkets/… request returns 404 EXCHANGE_NOT_AVAILABLE, router fan-outs exclude it, and it is not counted among the served venues. It returns when a commercial API agreement is in place. ↩