FIX Connectivity for Brokers, Funds and Trading Firms: Hosted Network, Licensed Engine or Build Your Own Gateway

FIX, short for Financial Information Exchange, is the standard message format that trading firms use to send orders and receive executions. FIX connectivity is the working link to each broker or venue: a session that stays logged in and recovers after a break without losing a message. You can rent that link from a hosted network, license a FIX engine and run it yourself, or build your own gateway. Whichever you choose, budget for people, testing with every counterparty and the hours someone must cover.

What is FIX, and who maintains it?

The FIX Trading Community calls itself “the custodian of the FIX Protocol”. It describes the standards as “open, technology-neutral specifications for electronic trading and trade processing”. The protocol “defines what business information is being exchanged and what it means”, and it covers orders, executions, quotes and market data across equities, fixed income, currencies and derivatives.

Two layers matter in practice. The application layer carries the business messages, such as a new order or an execution report. The session layer keeps the link alive. The FIX Trading Community's page on technical standards says the protocol “enables reliable, bidirectional communication between two parties through a session”, with ordered messages and continuous sequence numbers.

Which FIX versions matter in 2026?

The FIX Trading Community states: “Firms are strongly recommended to use FIX Latest.” It adds that, for firms with legacy implementations, “FIX Protocol versions 4.2 and 4.4 are still supported by the FIX Trading Community”, and that other versions are no longer supported. Your counterparties decide which version you speak. A broker that runs 4.2 will not move because you prefer something newer, so list the versions each counterparty offers before you pick any product.

What does a hosted FIX network or connectivity provider do?

You connect once to the provider, and the provider holds the sessions to the brokers, venues or clients on the other side. These are the providers' own descriptions, read on 10 November 2026.

• Broadridge offers to “Perform automated onboarding and instant certification of FIX counterparties through our web-based interface”, and says you can “Deploy as a fully or partially managed solution”.
• Iress describes its FIX Hub as “an end-to-end, fully managed service with low touch onboarding and FIX manipulation”, giving access to counterparties “via one connectivity partner, connection and certification”.
• Rapid Addition's RA Hub “Supports on-prem, in-cloud, or hybrid deployments”, so check who would hold the sessions. It says its counterparty connectivity offering aims to “Simplify and accelerate client onboarding, improving customer experience and reducing time to trade.”

Test each claim with your own counterparties. None of the three pages stated a price.

What do you give up with a hosted provider?

You give up control over the path your orders take and over the timing of changes. Your orders pass through someone else's servers. Iress mentions “FIX manipulation”, and Broadridge's managed service uses “a normalization service”. We read both as the provider changing messages in transit, though neither page says which fields or how. Ask exactly which fields it changes, and whether you see the original and the changed message side by side.

If the provider is down, you need another route to cancel what is working at the venue. Get that procedure in writing.

What does it take to run a licensed or open-source FIX engine?

An engine is a library that speaks the FIX session layer for you. OnixS describes its engines as “high speed implementations of the Financial Information Exchange (FIX) Protocol” in .NET, C++ and Java. They come with example code for “pluggable session state and message storage” and are offered on “a low cost fully supported subscription basis”. Its page shows no price, and offers a free 30-day evaluation. Open-source engines cost no licence fee. The QuickFIX/J page points to a mailing list and a Stack Overflow tag for questions. The QuickFIX project site also lists commercial support, so you either buy support or rely on your own staff.

The engine is the smaller part of the work. You still have to run these:

• Session management for each counterparty: connection times, credentials and sequence number resets.
• Message storage, so that after a restart you can answer a resend request, which is a counterparty asking for messages it missed.
• Certification with each venue or broker, repeated when a change may affect the connection or its messages.
• Monitoring that tells a person quickly when a session is down.
• Staff, or a rota, for every hour the markets you trade are open.

What is certification, and how long does it last?

Certification is a test set by the venue or broker before it lets your software trade live. ASX states: “All customer applications must pass the ASX Conformance process prior to accessing the Production market.” It requires your software to be “in a production-ready state, before booking the test”. It also reserves “the right to block access to the production system by non-compliant software”.

The test is not a one-off. ASX lists cases in which customers must repeat it, among them when software is “modified in any way that may affect or directly impact ASX connectivity or messaging”. Ask a hosted provider whether its earlier tests with a venue count for your orders, or whether you must run your own.

Where does the protocol end and your own work begin?

FIX is a standard, but every counterparty adds its own rules. The FIX Trading Community built FIX Orchestra because of this. It was “conceived as machine readable rules of engagement between counterparties”, meant to cut the time it takes to bring a new trading partner online. A rule book helps only when the counterparty publishes one. ASX and the London Metal Exchange publish theirs as PDF documents.

Two published venue and broker documents show how far apart the rules can be. Paxos, in its crypto brokerage documentation, states: “Cancel on Disconnect is enabled by default.” It also “forces a disconnect and reset of sequence numbers weekly for all sessions”. The London Metal Exchange specification differs on logout: “The gateway will not automatically cancel” a user's non-persisted orders. A user can ask for cancellation on connection loss with a flag on each order.

Your own work is order state: what your system believes about each order, and what you do when the counterparty disagrees. Each counterparty needs a mapping from its messages to your states, covering rejects, partial fills, cancels that arrive after a fill and amendments. An engine or a hosted network cannot write that for you, because it depends on your business rules.

When does latency matter?

It matters when your profit depends on being first to a price. For a broker routing client orders, or a fund working orders over hours, reliability usually matters more. OnixS describes its engines as “specially designed for mission critical, ultra low-latency, high frequency trading infrastructure”. That is the vendor's claim and its page gives no number.

Measure it yourself. Record when your system sends an order and when the matching execution report arrives, on a day of heavy trading. A hosted network puts the provider's systems in that path, and only your numbers show whether that matters to you. See also our articles on why milliseconds matter in trading and on exchange colocation.

What happens when a session drops in the middle of an order?

Four things can have happened. The order may never have reached the venue. It may still be working there. It may have executed, in whole or in part, with the report lost on the way back. Or the venue may have cancelled it on seeing the disconnect. Your system must find out which, without sending the order twice.

The session layer helps. For a gap during a session, the LME specification says, “The client may use the Resend Request (35=2) message to recover any lost messages.” At logon after a break it says gaps should be filled automatically, and “Neither side should generate a Resend Request (35=2) based on MsgSeqNum (34) of the incoming Logon message”. It also says its gateway “does not handle possible resends for the client-initiated messages”. It adds that “Any message with duplicate ClOrdID (11) will be rejected based on the Client Order ID uniqueness check”. So the gateway rejects a retry that repeats the identifier, and it processes one sent under a new identifier as a normal order. Find out what happened to the first order before you send it again under a new identifier.

Before you sign, run this test with the vendor: drop the connection with an order working, reconnect, and compare what each side believes. A team that builds should script it and repeat it after every release.

What duties apply when you connect to markets?

Rules differ by country and by what your firm is, so use this section as a map for the conversation with your lawyer. The quotations are not the whole of each regime.

• EU. Article 17 of MiFID II, the EU directive on markets in financial instruments, is published by ESMA, the EU markets regulator, in its online rulebook. It says an investment firm that engages in algorithmic trading “shall have in place effective systems and risk controls” so that its systems are “resilient and have sufficient capacity”. A firm using a high-frequency technique must store “accurate and time sequenced records of all its placed orders, including cancellations of orders”.
• EU testing. Article 6 of Delegated Regulation 2017/589, known as RTS 6 and written for investment firms engaged in algorithmic trading, says a firm “shall test the conformance of its algorithmic trading systems and trading algorithms” with the venue. That includes “when connecting to that trading venue through a sponsored access arrangement for the first time”. Sponsored access means sending orders directly to a venue under another firm's trading code, without using that firm's infrastructure. Article 12 requires the ability to “cancel immediately, as an emergency measure” unexecuted orders at all venues to which the firm is connected.
• US. SEC Rule 15c3-5 requires a broker-dealer with market access to “establish, document, and maintain a system of risk management controls and supervisory procedures”. The controls must be “reasonably designed” to “Prevent the entry of erroneous orders, by rejecting orders that exceed appropriate price or size parameters”. They must also be “under the direct and exclusive control of the broker or dealer”. One exception lets the firm allocate specific regulatory controls, by written contract, to a customer that is a registered broker-dealer. The chief executive “shall, on an annual basis, certify” them.
• FINRA. Regulatory Notice 15-09, published on 26 March 2015, lists suggested effective practices for firms that trade algorithmically. They include “testing is performed independently of code development” and “mechanisms by which the firm may quickly disable the algorithm or supporting platform”.

For the UK, an FCA review published on 21 August 2025 says algorithmic trading firms “are required to test the conformance of their algorithms with the system of the trading venue or direct market access provider, as per Article 6 of RTS 6”. It reports some firms noting that those procedures “differed significantly from venue to venue”. This article does not quote the FCA Handbook, so ask your lawyer about its chapters on algorithmic trading.

Depending on the regime and how the firm trades, these rules can mean a way to cancel everything at every venue, timed order records and evidence of testing. A hosted provider can supply parts of that.

When is building your own gateway justified?

Build when a requirement that earns you money is beyond every product you tested. Signs to look for:

• Your order state logic is the heart of your strategy, and a provider that edits or queues messages gets in its way.
• You need risk checks that run before each order, inside the same process that sends it, as in our article on the pre-trade risk hot path.
• The cost of a hosted network, counted over a few years across every session, exceeds the cost of a team to run your own.

Do not build because the engine licence looks costly. Certification, monitoring and the rota for market hours stay with you on every route. Our terminal article describes a general design for an owned FIX core.

What do you keep if you leave a provider?

That depends on the contract, so settle it before you start. Ask whether your certifications belong to the provider's sessions, so that a new route means new tests. Ask where the message history sits, and whose name each counterparty has on file for the credentials and network addresses.

Before you sign, ask for an export of every message in a readable format. Ask for the right to take over sessions with the counterparties' consent, and for a notice period long enough to retest. With an engine, you keep the sessions and the history, and with your own build you also own the code, if the contract says so.

What should you ask a vendor, or a firm that would build it?

• Which of our counterparties do you already connect to, and which FIX versions and certification dates can you show?
• Which fields do you change in transit, and do we see both versions of each message?
• What happens to our working orders if our session or yours drops, and can you show it live?
• Can our own staff cancel every order at every venue without calling you?
• Where is the message history kept, for how long, and in what format can we export it?
• Who repeats certification when we change our software or the venue changes its rules?
• Who is on call for each market's hours, and what is the response time in writing?

Where does amBrain fit?

amBrain publishes this blog. It describes part of its trading work this way: “amBrain builds algorithmic trading infrastructure: order execution, market data and pre-trade risk controls.” Its services card for trading and FinTech lists “Trading terminal development, order management systems, and FIX protocol exchange integration.” That is a service description only.

amBrain has been building software since 2019. On how it works with clients, it says: “Three formats: full delivery, a dedicated team, or engineers embedded in your team.” Its line on ownership is “The client keeps full ownership of the product and the code, except our reusable components.” Ask for the list of those components by name before you sign.

This article is not a case study and describes no client work. Nothing in it says that amBrain has built a FIX gateway, a connectivity product or a hosted network for any client. The quotations come from published pages of the FIX Trading Community, the vendors and venues named above, ESMA, FINRA, the FCA, the US Code of Federal Regulations and legislation.gov.uk. The article claims no relationship between amBrain and any of them, recommends none of their products and gives no amBrain prices or timelines.

It is not legal or investment advice either. Check the rules that apply to your firm with a lawyer in each market where you trade. If amBrain is on your shortlist, ask it the same questions as any other firm.