Letting an AI place trades on a real account is a reasonable thing to be cautious about. MTContext is built around that caution: by default, a trade request doesn't execute — it shows you exactly what would happen and waits for your explicit confirmation. This article explains the safety model so you know precisely what's happening before anything touches your account.
The default is a preview, not an order
Every trade-affecting request runs as a dry run unless you explicitly say otherwise. A dry run tells your AI client the full details of the order — symbol, direction, volume, price, stop loss, take profit — and validates it, but does not send it to your broker. Nothing executes until a request is made with dry_run: false deliberately.
In practice, that means when you ask your AI to "buy 0.1 lots of EURUSD," what comes back is a description of the trade and a check that it's valid — not a filled order. You review it, and only then confirm.
What gets checked before anything executes
Even when you do confirm a real trade, MTContext validates it first and fails safe if something's off:
- Lot size is checked against your broker's minimum, maximum, and step.
- Margin is checked against what's available in your account.
- Broker rules — symbol availability, trading hours, and restrictions — still apply exactly as your broker enforces them. MTContext can't override them.
If a check fails, the order is rejected rather than sent.
Other guardrails you control
- Equity protection — you can set a rule that closes positions if your account equity drops below a level you choose. See Setting up equity protection guardrails.
- Read-only by default on the free tier — the free plan exposes market and account data only, so there's no way to place a live trade at all until you're on a plan that includes trading.
- Every command is deduplicated — each carries a unique ID, so a repeated or retried request can't accidentally fire the same order twice.
What MTContext does not do
It doesn't trade on its own. MTContext is the bridge; your AI client does the reasoning, and you approve the actions. There's no background strategy placing trades while you're away unless you've explicitly set one up through your AI client and confirmed live execution.
A sensible way to start
Try read-only requests first (prices, account balance, open positions) — those never touch your account. When you move to trades, keep them as dry runs until you're comfortable with what the AI proposes, then confirm real execution one order at a time.
Next steps
See What MTContext can and can't do for the full boundary of what's possible, and Understanding the dry_run safety model for a deeper look. Ready to place your first real order? Do it from a connected client after reviewing the dry run.
Start free at mtcontext.com — the free tier is read-only, so it's a safe place to explore.