Everything that happens between the moment you click Buy in MetaTrader 5 and the instant your fill arrives. Server architecture, routing decisions, bridge software, the Ultency engine, and where latency actually comes from.
When you download MetaTrader 5 from your broker, you are installing a native client application. A Windows executable, macOS app, iOS or Android binary, or a WebAssembly build that runs in the browser. The client is identical across brokers because MetaQuotes licenses the same binary to every broker that runs MT5. What differs is which server it connects to and which account configuration sits behind that connection.
The client's responsibilities are narrow but specific:
Critically, the client does not connect directly to any exchange or liquidity venue. Every byte of market data and every order request flows through your broker's server cluster. This is why broker choice matters more than platform choice. MT5 is just the lens you look through.
What your broker actually licenses from MetaQuotes is a server-side platform that runs on Linux (Windows server support was deprecated in the MT5 Server 2300 update). A production MT5 server is rarely a single machine. It is a cluster of specialised processes:
| Component | Role | Typical hostname |
|---|---|---|
| Access Server | Authenticates logins, manages connections, load-balances clients | access1.broker.com |
| Trade Server | Receives orders, applies risk checks, routes to liquidity | trade1.broker.com |
| History Server | Stores tick data, trade archive, account statements | hist1.broker.com |
| Backup Servers | Hot standby for failover during maintenance or outages | access1-bk.broker.com |
When you log into MT5, you select a "server" from the dropdown. That string (e.g. ICMarkets-Live01) is your access server's identifier. The access server then directs your terminal to the appropriate trade and history servers for your account group.
MT5 uses a proprietary protocol over TCP/IP, defaulting to port 443. The same port as HTTPS. This means MT5 connections generally pass through corporate firewalls without special configuration. As of MT5 Build 5660 (February 2026), HTTP and SOCKS5 proxy support has been enhanced and the older SOCKS4 protocol is no longer supported.
Once your order arrives at the broker's trade server, it runs a sequence of checks:
The routing decision is configured per-symbol and per-account-group by the broker's dealing desk. There are three common models:
| Model | Behaviour | Conflict of interest |
|---|---|---|
| Full A-book (STP / DMA) | Every order hedged externally to an LP. Broker earns from spread markup or commission. | None. Broker profits whether you win or lose. |
| Full B-book (Market Maker) | Broker takes the opposite side internally. Broker profits when traders lose. | Structural. The broker is on the other side of your trade. |
| Hybrid (most common) | Profitable traders and large positions go A-book. Small lots and statistically losing accounts stay B-book. | Partial. The broker chooses which trades to internalise. |
None of this is visible in the MT5 terminal by default. You see a fill price, a timestamp, and sometimes a trade comment. The routing model is inferred from execution behaviour, not displayed.
Until 2024, brokers running A-book had a problem. MT5's native architecture was not designed to talk directly to external liquidity providers. The solution was a category of middleware called bridge software. A separate process that sits between the MT5 trade server and the broker's FIX connections to LPs.
The three dominant bridge vendors:
Bridges add complexity and cost. They also add 5 to 30ms of latency depending on deployment topology. This is the problem Ultency was built to solve.
Launched in 2024 and rapidly adopted in 2025 to 2026, Ultency is MetaQuotes' native matching engine that runs inside the MT5 server stack. Instead of running a third-party bridge, brokers can connect to liquidity providers directly through Ultency's FIX gateway.
The technical pitch:
Several institutional brokers have announced Ultency integrations in the past 12 months. For traders, the practical impact is faster fills and tighter spreads on A-booked flow. Whether your broker uses Ultency is rarely advertised on their website. You would need to check broker disclosure documents or ask support directly.
"Low latency" gets thrown around in broker marketing. Here is where the time actually goes when you place a market order:
| Stage | Typical latency | Who controls it |
|---|---|---|
| Click to client sends request | 1-5 ms | Your machine and MT5 client |
| Your ISP to broker data centre | 10-200 ms | Your internet (biggest variable) |
| MT5 trade server processing | 0.5-2 ms | Broker infrastructure |
| Bridge processing (if used) | 5-30 ms | Broker bridge config |
| Ultency processing (if used) | 0.1-1 ms | MetaQuotes infrastructure |
| LP venue execution | 1-10 ms | Liquidity provider |
| Fill confirmation back to you | 10-200 ms | Reverse of above |
The single biggest factor for most retail traders is the network hop between your ISP and the broker's data centre. Brokers typically host MT5 servers in Equinix LD4 (London), NY4 (New York), or TY3 (Tokyo). The same facilities that host major institutional venues. If you trade from a country far from those hubs, latency from your home connection dominates everything else combined.
Traders serious about execution speed run MT5 on a VPS co-located in the same data centre as their broker's server. This reduces round-trip from hundreds of milliseconds to single digits.
You can never know for certain. But there are signals.
On A-booked execution, some brokers leave the LP name or venue code in the trade comment field. If you see strings like LMAX, UBSEX, CITIFX, that is a strong A-book signal. If comments are blank or generic ("market"), it is inconclusive.
A-book brokers tend to show symmetric slippage. Sometimes you get a better price than expected, sometimes worse. B-book brokers often show asymmetric slippage. Losses are slippage-padded but gains are exact. Track 100 plus market orders and run a histogram.
A-book spreads widen visibly during news events because the LP's spread widens. B-book brokers can keep spreads artificially stable but may issue requotes or trade rejections during the same windows.
Tier-1 regulated brokers (FCA, ASIC, CySEC) publish execution quality reports under MiFID II or similar frameworks. Look for the broker's "Order Execution Policy" and "Best Execution" disclosures. They will often describe their hybrid routing in general terms.
MT5 defaults to port 443 (the same as HTTPS), which means it generally passes through corporate firewalls without configuration. Some brokers offer port 80 as a fallback. The connection is encrypted using a proprietary protocol over TCP/IP.
No. Your EA runs locally on your MT5 client (or your VPS). The broker only sees the orders it produces. Same as if you placed them manually. The EA's source code, parameters, and decision logic never leave your machine.
Yes. Network latency depends on physical distance. A server closer to you (or your VPS) gives faster fills. Data centre matters for liquidity. Brokers in Equinix LD4, NY4, or TY3 are typically co-located with their LPs, reducing LP round-trip latency.
Ultency is offered by MetaQuotes to MT5 broker licensees, but uptake varies. Larger institutional brokers have been faster to integrate. Smaller retail brokers often stick with their existing bridge software. There is no public registry of which brokers use Ultency. You would need to ask the broker directly.