A terminal that trades several asset classes needs access to each market, data licences for the exchange prices it shows or uses, and one master record for every instrument it offers.
If you have funding for a multi-asset trading terminal and no engineers of your own, talk to engineering firms whose trading systems already run in production for clients, connected to real brokers and exchanges. Each asset class you add brings more connections and data licences, and many of the parts are usually rented rather than built. Six questions further down separate firms that have built trading systems from general outsourcing companies.
The short answer: shortlist firms that can show a trading system they built running in production today, ideally in more than one asset class. Before the first call, list the brokers and exchanges you plan to use and where your prices will come from. Then put the six questions below to every firm and compare the answers.
Read next
A terminal for one market may need only one broker or exchange. Each asset class you add, whether equities, futures, currencies (FX) or crypto, usually brings more connections and data licences, plus its own instrument rules and trading hours. If your company will take orders from clients, a new asset class can also change which regulatory licence it needs. Ask a regulatory lawyer about each market before the build starts.
The extra work falls into five parts.
Each venue, meaning an exchange or another trading platform, needs its own connection, and so does each broker. The connection carries your orders out and brings back prices and fills, the confirmations that an order has traded. Each one has its own interface and its own rules for what happens when the line drops.
Exchange prices, known as market data, come with licence terms set by the exchange, even when a broker or a data vendor delivers them. Each exchange whose prices you show adds its own agreement.
At many exchanges, use of prices by software alone counts as a use of its own. Nasdaq calls it “non-display use”. Its Data News #2015-9 of 14 December 2015 defines the term as “any method of accessing Nasdaq U.S. Information that involves access or use by a machine or automated Device without access or use of a Display by a natural person or persons”. The notice gives examples of uses that carry a fee, including “All automated trading programs, applications, and scripts” and risk management with automatic stop-loss functions. If your software reads prices to trigger orders or risk checks, ask each exchange what that requires.
An instrument is anything a client can trade, such as a share, a futures contract, a currency pair or a coin. The instrument master keeps one record for each instrument, with the names every venue uses for it. The record also holds the currency and the contract size, and for futures the expiry date.
Two public standards help. The Market Identifier Code (MIC, standard ISO 10383) is a four-character code for each exchange or trading platform, such as XNYS for the New York Stock Exchange. The Financial Instrument Global Identifier (FIGI) is a code for the instrument itself.
The MIC page on iso20022.org says the ISO 10383 standard specifies “a universal method of identifying exchanges, trading platforms, regulated or non-regulated markets and trade reporting facilities as sources of prices and related information”. SWIFT keeps the official list on behalf of ISO and publishes updates every month.
OpenFIGI, the public site for the identifier, says Bloomberg is the body that registers the codes. It describes a FIGI as “a 12 character, alphanumeric, randomly generated ID”, issued under a standard of the Object Management Group, a non-profit standards body. FIGI covers asset classes that usually lack a global identifier, including “cryptocurrencies and pairs on exchanges, futures and options”. A FIGI does not change once it is assigned.
Markets keep different trading hours. On the New York Stock Exchange, the core trading session for shares runs from 9:30 a.m. to 4:00 p.m. Eastern Time, opening and closing with an auction, according to NYSE's trading hours page. Kraken, a crypto exchange, writes in its guide “What makes crypto 24/7/365?” that “the crypto market never sleeps”. One screen has to show which markets are open now, and what happens to an order a client places while its market is closed.
Order types differ as well, since one venue may reject an order type that another accepts. The terminal should offer each client only the orders the chosen venue will take at that moment.
A multi-asset terminal also needs one view of positions and risk across accounts. A client may hold shares and futures through two brokers and crypto on an exchange. Totalling them needs live prices from every source and one rule for converting currencies. The view must also show plainly when a source is late or down, so nobody acts on a stale total.
Few terminals are built from scratch. A typical split looks like this.
Usually rented or licensed:
Usually built:
If clients will pick you for the screen, build the screen and rent most of the rest. If they will pick you for access to a market few others reach, build and own the connection to that market.
Talk to engineering firms that build trading systems for clients and can show one running in production today. Such a firm should be able to tell you which venue tests its systems passed and how the data licences were handled.
General outsourcing companies can supply many developers, and some have a trading group inside. If you talk to one, meet the engineers who would build your terminal and ask each about a live trading system they worked on.
A firm that has built a terminal for one crypto exchange knows only part of the work. Ask what it has done with brokers, futures or shares, where connection tests and data licences work differently.
Put the same six questions to every firm on your list.
Which asset classes have your systems traded in production, and through which brokers or exchanges? A firm with real experience names them and says which of those systems still run today. Ask whether any single system handled two asset classes at once, and what had to change when the second one was added.
How do you keep one record for each instrument? Listen for venue codes such as the MIC and an identifier such as FIGI or a data vendor's own. A good answer also covers futures expiries and share splits. If the answer is that the ticker is enough, the firm has not met the problem yet.
How did you get through each venue's connection tests? Many brokers and exchanges test new software before they let it trade. A terminal with several venues can face several rounds of tests, and each venue sets its own test dates. Ask what failed the first time and what had to change before the venue approved it.
What does a client see when one venue goes down? Prices from that venue are marked as stale and new orders to it are blocked, while the other markets keep trading. When the venue comes back, positions are checked against the venue's own records before trading resumes.
Which latency figures did you measure on a live system, and between which two points? Latency is a delay inside the system, for example from a price arriving to an order going out. A figure without its two end points and a note of how busy the system was cannot be compared with another firm's figure or with what your clients need.
What happens after launch, and what will we own? If one of your markets never closes, there is no quiet night for updates. Ask when the firm releases changes and who answers alerts at weekends. Then ask whether the venue adapters and the instrument master become yours or stay with the firm as its own components, and on what licence you keep using anything it keeps.
amBrain has been building software since 2019. amBrain builds algorithmic trading infrastructure: order execution, market data and pre-trade risk controls.
One line on amBrain's website reads: “Trading terminal development, order management systems, and FIX protocol exchange integration.”
One of amBrain's measured figures is <5 ms market data latency. Another is <1 ms risk latency.
amBrain describes its team in one line: “A team of up to 40 people, about 75% of them senior.” It works in three formats: full delivery, a dedicated team, or engineers embedded in your team. The client keeps full ownership of the product and the code, except amBrain's reusable components.
Nothing above describes a specific amBrain project or client, and the article names no prices or timelines.
If amBrain is on your shortlist, send it the same list of venues and asset classes, and the same six questions, that you send every other firm.
Bring your current architecture and the failure mode that worries you, and we will go through it together in half an hour.