About PredictionMarkets.tools

A catalogue of the venues, data and software people use to trade and forecast events — who may use it, what it settles in, what a fill costs and who decides the outcome, in the same blocks on every card.

What is catalogued

5 categories, each opened once there was enough in it to be worth reading end to end:

Every card in every category also sits in one flat, filterable list of all tools.

A category answers “what is this bought for”. The collections cut the same catalogue the other way, by a property that runs across every category — with an mcp server, open source, play money, cross-venue, streaming feed, self-hosted, calibration scoring, cash-settled — so a reader who has already decided how a tool must run can start there instead. A card’s home stays its category; a collection is a selection, not a second address.

Some products are deliberately not catalogued at all, and what to use instead is for the reader holding one of them: 12 pages on products that shut down, priced themselves out of reach, or closed the terms you were building on, each listing what in here replaces it.

And some of what stops a reader is not a product at all. The guides are 20 reference pages on how this market works — who decides how a contract resolves, why a venue will not serve your state, what a trade actually costs once the fee is converted — each claim carrying the document it came from. They are corrected in place rather than dated, and they recommend nothing.

Where the boundary runs

By product type, not audience. Inside: venues, forecasting platforms, data, APIs, libraries and the clients that trade through them. Outside: tipsters, signal sellers, odds aggregators, and anything whose product is telling you which side to take. That is a deliberate choice to stay out of advice and stay in tools.

The unit is a product, not a company. One vendor can appear more than once — a web terminal and the API behind it have different prices, different buyers and different competitors, so each gets its own card rather than one entry that describes neither.

How a card is made

Every card fills in the same blocks, whatever its category, and each block carries its own verification date — pricing goes stale in weeks, availability overnight. A field nobody checked is left blank rather than guessed.

What none of it rests on is having traded here. Tested hands-on is a badge a card earns when we have held an account or a key and used the thing, and none of the 60 cards carries it: funding accounts across venues in this sector is a cost this catalogue has not taken on. So what is offered instead is narrower, and checkable — every claim read out of a primary document and dated, so you can see how old it is and go and disagree with it.

The methodology page covers how sponsorship and affiliate links are handled; the affiliate disclosure covers the links themselves.

Three catalogues, one author

The same person builds two more catalogues on the same plan, for two markets this one does not cover: StockMarketStack for stock-market software and data, and CryptoMarkets.tools for crypto trading and market data.

They are separate sites rather than sections of one because the reader is different. What decides a question here is who resolves a market, whether a venue will serve you at all and what a trade actually costs; on the other two it is exchange fees, funding rates and data licences. A card written to serve both would answer neither well.

Where the same product is bought for two of those markets, the cards say so and link across: the CCXT card here is about its prediction namespace, and points at the crypto one for the 104 exchanges that are not this catalogue's subject. Nothing else is shared between the three sites except the software that renders them.

Nothing here is investment advice — this site describes tools, not what to do with them.