99xi.com

MetaTrader 5 is software, not a broker. That single fact shapes every decision below, including the one most guides get wrong: there is no financial regulator that licenses a "forex VPS," and nothing about hosting your terminal changes who holds your money or what protections you have.

What hosting does change is uptime, and whether your Expert Advisors keep running when your laptop is shut. This piece covers the three routes to that, the hard limits MetaQuotes publishes in its own service rules, how to pick a data centre using turnover data instead of guesswork, and the checks worth doing before you pay anyone.

Three ways to run MT5 around the clock

MetaQuotes sells a virtual hosting service directly inside the terminal. You can also rent a Windows VPS from any hosting company and install MT5 yourself. Or your broker may supply one, usually free if you clear a balance or volume threshold set out in the client agreement.

Which one suits you depends less on price than on what your EAs need. MetaQuotes' own service is the cheapest to start and the most restrictive by design. A third-party Windows VPS gives you a full desktop and lets you pick the city. A broker-supplied one costs nothing but locks you to that broker and usually comes with modest specifications.

What MetaQuotes' own hosting rules forbid

These are not marketing caveats. They are written into the Rules of Using the Virtual Hosting Service on mql5.com, and they decide whether the built-in option is usable for you at all.

  • No DLLs, at all. The rules state there is no physical capacity to use them, and if a program on a virtual terminal tries to call a function from any DLL, that EX4 or EX5 program stops immediately with a critical error. Any EA that imports a DLL is disqualified from this route outright.
  • One trading account per rented terminal. The same account may run on several virtual terminals, but one terminal cannot serve two accounts.
  • 32 charts on paid hosting, 16 on free. The same cap applies to custom Expert Advisors launched on charts. Charts beyond the limit are simply not migrated, and you get a journal message instead.
  • Maximum 500,000 bars per chart. Whatever you set in the terminal's "Max bars" field, migration caps it at that figure.
  • No physical access. You get logs and load monitoring through your own client terminal. There is no desktop, no file system, and no way to run anything other than the terminal.
  • Synchronisation is manual. The virtual terminal's environment does not update itself. Change a chart or an EA locally and you must synchronise again for it to take effect in the cloud.
  • Scripts, non-standard timeframes and chartless symbols are not migrated. Neither are charts with no EA attached, which surprises people the first time.
  • Maintenance happens on someone else's schedule. A hoster may take servers down for up to 4 hours a day at weekends and up to 15 minutes on weekdays.
  • Minimum rental is one month and cancelling gets you nothing back. The rules say paid rent cancelled by a user is not refunded, and rent terminated because your EAs disturbed the hoster's hardware is also not refunded.
  • One rule works in your favour: automated trading is always enabled on a virtual terminal, even if you disabled it in the terminal settings or in the EA's parameters.

    Choosing between the three routes

    Prices for third-party and broker-supplied hosting change constantly, so the table compares constraints rather than monthly fees. Those stay true after the pricing pages move.

    MetaQuotes virtual hostingThird-party Windows VPSBroker-supplied VPS
    DLL-based EAsBlocked. Critical error on any DLL callAllowed, subject to your Windows licenceUsually allowed, check the client agreement
    Accounts per terminalOneAs many instances as the machine carriesSet by the broker
    Chart and EA cap32 paid, 16 freeLimited by RAM and CPUUsually modest
    Choose the data centreAssigned from the hosting cloud networkYou choose the cityFixed by the broker
    Desktop or file accessNone, logs onlyFull RDP desktopVaries
    Software other than MT5NoYesRarely
    Stated maintenance windowUp to 4 h at weekends, 15 min weekdaysPer the host's SLAPer the broker
    Minimum term and refunds1 month, no refund on user cancellationOften hourly or monthly, check termsTied to the balance or volume condition
    Published price15 USD per month at 1 month, 13 at 3, 10.83 at 6, 10 at 12, per mql5.comVaries by provider, verify on their siteUsually free if you meet the stated condition

    The MetaQuotes figures above were read from mql5.com and are the vendor's current published rates. Confirm them on the day you order, because they can change. There is a 24-hour free trial, and the service rules state that you are switched to the paid plan automatically once it ends, so set a reminder if you intend to test and walk away.

    Picking a data centre from turnover data

    "Choose a server near your broker" is the standard advice, and it is right for a reason you can actually check. The BIS Triennial Central Bank Survey, published on 30 September 2025, puts global OTC FX turnover at $9.6 trillion per day in April 2025, up 28% from 2022. Sales desks in four jurisdictions accounted for 75% of that activity: the United Kingdom, the United States, Singapore and Hong Kong SAR. The UK alone held 38% of total turnover, and Singapore reached 11.8%.

    Where sales desks sit is where liquidity providers and prime brokers place their matching engines. That is why the same handful of facilities keep coming up in this niche: Equinix LD4 in London, NY4 and NY5 in New York, TY3 in Tokyo, and the Singapore sites. If you do not know where your broker's trade server physically is, ask support for the hostname or the facility, then resolve it and trace the route.

    Cross-check the answer against the pair you actually trade. A broker running Asian-session pairs out of a Tokyo facility and a broker running USD crosses out of London will point you to different cities, and copying someone else's choice helps only if you share their broker.

    How to measure your own latency instead of trusting a vendor

    I am not going to give you a millisecond figure, because any number that does not name both endpoints is decoration. Latency is a property of one specific path: the machine running your terminal, and the one specific server your broker matched your order against. Change either and the number changes.

    Measure it yourself, like this.

  • Get the trade server hostname from your broker, then run a standard round-trip measurement against it from your candidate VPS and separately from your home connection.
  • Sample across a full session rather than once. A five-second reading at 03:00 server time tells you nothing about the London and New York overlap, which is when your EAs are busiest.
  • Read the ping indicator your own terminal displays for the connected server, and use it as the cross-check against your network measurement.
  • Repeat over several days before you conclude anything. Single samples are noise.
  • Compare against the alternative, not against an ideal. The useful question is whether the VPS beats your home connection to the same server, by how much, and whether that gap is worth the monthly cost for your strategy's holding times.
  • Vendor claims are vendor claims. MetaQuotes publishes that its optimised core gives 0-5 ms latency against 100 ms from a home PC, that 96% of broker servers can be reached in under 10 ms and 84% in under 3 ms, that it operates more than thirty hosting points, and that hardware is specified for up to 99.99% uptime with up to 3 GB of RAM and 16 GB of disk per virtual machine. Those are the provider's own published statements, not independent measurements, and you should read every competing provider's numbers the same way.

    The compliance boundary in this niche

    Hosting infrastructure is not a regulated financial activity. No authority issues a forex VPS licence, and a provider that implies otherwise is selling you something. The real constraints sit in three places, and the first one is worth reading in the vendor's own words.

    The MetaQuotes service rules state that the administration does not own the physical servers in its hosting cloud network and makes no guarantee of smooth operation. Section VII says the service takes no responsibility for losses or short-received profit from using or being unable to use it, and that MetaQuotes is a technology provider which gives no financial advice, with all transactions processed through your broker. The same section lists the risks outright: hoster hardware malfunction, MQL4 or MQL5 program errors sending wrong or late orders, simultaneous connections from different places on one account, and communication failures causing trade requests to go unfilled.

    The second constraint is your broker's client agreement. Some permit Expert Advisors and some restrict them. Several prohibit strategies that exploit price feed latency or that place and cancel orders at a rate their infrastructure was not built for, and those clauses exist whether or not your EA does so deliberately. Read that section before you deploy, because the sanction is a closed account rather than a refund.

    The third is data location. You are putting credentials, EA logic and trade history on someone else's machine, often in another country. That has privacy and, for some traders, reporting consequences that no hosting review can answer for you.

    Checks before you pay

  • Ask your broker whether EAs are permitted at all, and whether VPS hosting triggers any condition in the client agreement.
  • Confirm the refund and cancellation terms in writing before the first payment, and note whether the minimum term is a month or an hour.
  • Find the stated maintenance window and decide whether it collides with your trading hours.
  • Check whether the provider takes backups, and whether restoring one is self-service.
  • Decide whether you need one account or several. If several, MetaQuotes hosting means one rented terminal each.
  • Test with the minimum viable specification and the shortest term, measure for a week, then commit. Upgrading later costs far less than a year of the wrong plan.
  • The failure mode most people hit is not latency. It is discovering after a month that the environment cannot run what they already built, which is why the DLL rule and the chart cap belong at the top of your checklist rather than at the bottom.

    This article describes published vendor rules and publicly available market statistics. It is not investment advice, and nothing here creates any regulatory protection for funds held with a broker. Confirm all pricing and terms on the provider's own site before committing money.