An algo doesn’t get tired, doesn’t tilt, and doesn’t skip the plan — which is exactly why some prop firms welcome it and others ban it outright. Before you deploy a single EA on funded capital, find out which camp your firm is in, because a bot that breaks a rule breaks it thousands of times before you notice.
Which firms allow automation — and which don’t
There is no universal answer, so sort firms into three buckets and confirm which one yours sits in:
- Fully allowed — EAs, custom bots, and copiers are permitted with normal risk rules applying. Common among forex-CFD firms built on MetaTrader.
- Conditionally allowed — automation is fine except specific prohibited techniques (below), or is allowed on evaluation but restricted when funded.
- Banned — manual discretionary trading only, sometimes detected by behavioral flags on order timing and uniformity.
The word “algo” means nothing to a rulebook. What matters is the list of prohibited techniques, which most firms police regardless of whether you press the button or a script does.
HFT, latency-arbitrage, and gap-exploit bans explained
Almost every firm — even the automation-friendly ones — bans a specific family of strategies because they exploit the broker’s pricing rather than a real market edge. If your bot does any of these, it will be flagged no matter how profitable:
| Prohibited technique | What it exploits | Why firms ban it |
|---|---|---|
| Latency arbitrage | A pricing feed that lags the real market | Profit comes from the broker’s delay, not the trade |
| Tick / HFT scalping | Sub-second fills on micro-moves | Order rate and fill quality strain the model |
| Gap / weekend exploits | Predictable Monday-open gaps | Bypasses genuine risk-taking |
| Reverse-arbitrage & copy exploits | Coordinated fills across accounts | Manufactures results the firm must pay on |
Read the prohibited strategies section as carefully as the drawdown rules. A legitimate trend-following EA is usually fine; a bot whose entire edge is the broker’s feed will get a payout denied even when the P&L is real. When your strategy sits near a gray line, ask support in writing before you fund.
Firms friendly to automation
Some firms are explicitly comfortable with EAs and bots:
- FTMO — permits EAs and automated strategies within its rules, with the standard carve-outs for latency arbitrage and other exploit classes. A long-standing choice for MetaTrader-based automation.
- FunderPro — positions itself as automation-friendly and supports common retail platforms, making EA deployment straightforward.
As always, terms move. Confirm on the day you buy that your specific account type allows automation on both evaluation and funded stages, and that your technique isn’t on the banned list.
Platform and API support for your algo
Your bot is only as good as the platform it runs on. Two practical checks decide whether deployment is smooth or a fight:
- MetaTrader 5 is the workhorse for EA traders — mature MQL5 ecosystem, huge library of existing strategies, and broad prop-firm support. If your algo is written for MT5, prioritize firms that offer it natively.
- cTrader appeals to traders who want low-latency execution, depth-of-market, and a cleaner automation API (cBots / Open API). If your strategy is latency-sensitive but legitimate, the execution layer is part of the edge.
Also verify the firm doesn’t block VPS hosting or automated order routing in the fine print — an EA that must run on a colocated VPS is useless at a firm that forbids it.
Auto-journaling algo trades to catch edge decay live
Algos fail silently. A strategy that printed money in backtest can quietly stop working live — regime shifts, spread creep, a data source that changed — and by the time the equity curve makes it obvious, you’ve given back weeks. The only defense is measuring the live edge continuously, not reviewing it monthly.
That measurement is tedious to do by hand across hundreds of automated fills, which is the whole point of automating the journal too. Shibiki auto-journals every fill the instant it closes, so an EA doing forty trades a day logs itself, and each strategy carries a live edge-health score with a Wilson confidence interval — the right statistic for high trade counts, because it separates a genuine drop in edge from ordinary variance. When the interval starts drifting below your calibration, you get told while it matters, not after. Pipe enough live trades through the expectancy calculator view and the question “is my bot still working?” becomes a number instead of a hunch.
Enforce risk limits at the broker your algo can’t override
The nightmare scenario with automation is a bug — a flipped sign, a runaway loop, a stale price — that fires orders faster than you can pull the plug and drains the account past the firm’s line. A stop-loss coded inside the bot is no protection when the bug is in the bot.
The fix is a limit that lives outside the algo, enforced at the broker: a hard daily-loss and max-position ceiling set a margin inside the firm’s rule, that flattens and blocks new orders regardless of what the script is trying to do. Shibiki holds that line at the broker layer, so a misbehaving EA hits a wall it cannot code around, and if you run the same algo across several funded accounts you can copy it — and the hard limit — across all of them from one master. Let the bot trade; keep the kill-switch in something the bot can’t touch.
Related: FTMO · MT5 integration · expectancy calculator