Where forex EAs are sold
MQL5 Market. The largest channel: thousands of MT4/MT5 EAs, bought inside the terminal, protected by activations. Reviews are visible but easy to inflate; the “signal” link on a product page is the only verifiable part.
Vendor websites. Established developers such as Profalgo, FX Automater or Forex Fury sell directly, often with their own Myfxbook accounts and set-file support. Licences are usually account-number bound rather than activation bound.
Marketplaces with live verification. AlgoTradingSpace lists EAs together with a calibrated live account that the marketplace itself runs, plus a published max floating loss. That is the closest thing to an independent test before purchase, and the reason our EA dashboard tracks every EA sold there.
Telegram, Discord and “EA cracked” sites. Not a channel. Cracked EAs are modified binaries with no support, no updates and sometimes a lot-size multiplier you did not ask for.
Licence types and what they mean for you
| Licence model | Typical channel | What you get | Watch out for |
|---|---|---|---|
| Activations | MQL5 Market | 5–10 installs, hardware-bound | VPS migrations burn activations |
| Account-bound | Vendor sites | Unlimited installs on listed account numbers | Changing broker means asking the vendor |
| Rental | MQL5 Market | 1–12 months, same activations | Total rental cost overtakes purchase within a year |
| Subscription | Some vendors, marketplaces | Monthly fee, updates included | Strategy changes mid-subscription without notice |
Prices: what is normal, what is a warning sign
From the EAs we track, one-time prices cluster in three bands: 100–500 USD for single-market scalpers and trend robots, 600–1,000 USD for EAs with multi-year public signals, and 1,500–2,000+ USD for best-sellers with stair pricing, where the vendor raises the price with every ten copies sold. Price says nothing about quality. Quantum Queen at 1,999 USD scores 3.4 in our test; Dark Gold EA at a fraction of that scores 4.0.
Two price patterns should make you pause. “Price rises to X after the next 10 copies” is a sales device, not a scarcity signal. And a lifetime licence under 50 USD for an EA that claims 30 % a month means the vendor earns from volume, not from traders who stay.
The red flags we see most often
- Backtest only, no live account. Ninety-nine per cent modelling quality in the strategy tester is not a track record.
- Live account younger than six months. Gold and indices have produced six-month windows where a martingale looked flawless.
- “No stop-loss needed.” Every EA without a stop-loss has one; it is your account balance.
- Drawdown quoted from the backtest. Ask for the live figure. Gold Trade Pro advertises about 10 %; the live account shows 26.7 %.
- Set files only in the private Telegram. Support that lives outside the product page disappears with the vendor.
- Vendor with 20+ products and a new release every month. Each release resets the track record.
Our why EAs fail live guide walks through the mechanics behind each of these.
Checklist before you pay
- Open the EA’s live account (Myfxbook, FX Blue, MQL5 signal or ATS account) and note age, gain, maximum drawdown and profit factor.
- Read the strategy description for the words grid, martingale, averaging, recovery. If they are there, read our grid and martingale risk guide first.
- Check the licence: activations or account-bound, and how many.
- Check whether your broker is EA-friendly with raw spreads — the ECN brokers list shows which are.
- Look up the EA in our reviews. If we run it live, the page links the account.
- Decide the lot size before buying, from your risk per trade, not from the vendor’s default. The position-sizing guide has the formula.
After the purchase
Install on a VPS, start with half the vendor’s suggested lot size, and put the account on a tracker from day one so drawdown and floating loss are visible without opening the terminal. Our free portfolio tracker does this for any number of MT4/MT5 accounts. Give the EA three months and at least one news week before you judge it — and compare its live curve against the vendor’s signal, not against the backtest.