One feed, six venues, one search box. The aggregator's job is to make a fragmented asset class feel like a single market, and it does that job well enough that the operational reality arrives as a surprise several weeks into the project.
There is no single market. There are six independent counterparties, each with its own legal entity, its own jurisdiction, its own account types and its own view on whether a fund is a category of customer it wants. Onboarding is six projects, and the useful first move is working out which of them are projects at all.
Read the coverage taxonomy as a triage list
Prediction Alpha describes its own coverage in six lines: on-chain permissionless markets, CFTC-regulated US event markets, play-money community markets, academic real-money political markets, a forecasting community with calibrated probabilities, and sports and entertainment prediction markets. Six venues, six descriptions, and they are not six of the same thing.
Two of those cannot be execution venues under any legal analysis, because there is nothing to execute. A play-money venue produces no fill in money and no settlement in money. A forecasting community publishing calibrated probabilities produces a forecast, which is an input to your model rather than a destination for your order. Neither belongs in the onboarding queue at all, and both belong in the research source register with a written line saying why they were excluded. An unwritten exclusion looks like an oversight when somebody reviews the file later.
That triage costs you an afternoon and removes a third of the workstream before a single document has been requested. Do it first.
What I am not going to tell you, and why
The obvious thing to want from an article with this title is a list: these venues will take a regulated US fund, these will not. I am not going to give you one, and you should be suspicious of anybody who does.
Eligibility is set by each venue's terms of use and by its own regulatory permissions. Both change, sometimes quickly, and neither is visible on any aggregator screen, including the one above. A list published today has no date attached by the time it is read, and the failure mode is not that you miss an opportunity. It is that you spend six weeks of legal and operations time onboarding to a venue on the strength of a stale claim, then discover the constraint at the account opening stage.
What you do instead is unglamorous and works. Put the eligibility question to each venue in writing, naming your entity type and jurisdiction explicitly rather than asking in general terms. Get the answer in writing. File it with the date you received it and the name of the person who sent it. That file is what your compliance function needs and what a stale list can never be.
The entity pack, and the trap that is specific to entities
The document set is roughly what you would expect: formation documents, jurisdiction and registered address, beneficial ownership above whatever threshold applies, a list of authorised traders with their scope, source of funds evidence, tax forms, and a funding rail in the entity's name rather than anyone's personal name.
The trap is not in that list. It is in the account type, and it kills more venue relationships than the legal question does. Many venues in this asset class were built retail first and have exactly one account type, which is a natural person's account. An account that is legally an individual's, funded by a fund and traded by employees, is a compliance problem regardless of how comfortable everyone feels about it. It is also the arrangement that desks drift into, because an individual account can be opened in an afternoon and the entity conversation takes a month.
So ask four questions before anything else, and ask them by name. Does an entity or institutional account type exist. Can the account be titled in the fund's name. Can the settlement asset be held consistently with our custody arrangement. And what are the per account position and order limits, because a cap designed for a retail customer is a capacity constraint on your strategy that no order book will ever show you.
Where custody and settlement break the standard model
Execution in this asset class usually runs through your own wallet or a trade-only API key rather than through a broker. Three consequences follow, and each of them is a question for your administrator and auditor before the account is opened, not after.
The venue is your direct counterparty. There is no intermediating broker absorbing settlement risk, so venue credit exposure is a real line in the operational risk register rather than a theoretical one.
Collateral is resident at the venue from entry until resolution. On a contract ending in January 2027, that is a residency period of more than a year during which the capital is neither deployed elsewhere nor held by your custodian.
Reconciliation runs against a venue balance rather than a custodian statement. Your administrator has to be willing and able to price the instrument and reconcile to that source. If they are not, the project ends there, and it is much cheaper to find that out in week one than in week seven.
Sequencing so a refusal costs a week rather than a quarter
Most desks run this workstream backwards, starting with KYC because it is the visible part and discovering the disqualifiers afterwards. Order it by cheapest disqualifier first.
Does an entity account type exist at all. Will the venue accept our jurisdiction and entity form, in writing. Can our administrator price and reconcile the instrument. Will our auditor accept the custody arrangement. Only then, full KYC and funding. Each of the first four is a conversation measured in days and any one of them can end the project, so paying for them before the expensive step is straightforwardly correct.
Two more things belong in the plan. A venue that cannot be onboarded is still a data source, and it should be admitted as one explicitly, with a written exclusion from routing so that nobody later mistakes an operational failure for an oversight. And the workstream ends with a number the portfolio manager needs before research spends anything further: if you finish with two tradeable venues and a per account cap on one of them, then the strategy's capacity was set by operations rather than by the opportunity. Send that number upstream early. It is a far better conversation to have before the research budget is committed than after the first allocation meeting has already assumed a size nobody can implement.