An SMC trade is four decisions, not one. Three of them automate cleanly and one of them does not. Which is which, why the order matters, and how a TradingView alert reaches a live account without the chart becoming the trader.
Count the decisions in one Smart Money Concepts trade. There is the read — is this leg impulsive, has structure actually broken, is the imbalance the kind that gets filled. There is the level — which order block, which edge of it. There is the trigger — do you take the touch, or do you wait for a lower-timeframe shift. And there is the size — how much, given where the invalidation sits.
Four decisions. Here is the answer to the headline question up front: three of those four automate cleanly, and the first one mostly does not. The read is where a discretionary SMC trader's edge actually lives, and every attempt to encode it into a rule has to survive the fact that the rule will fire on the chart that looks right and loses as eagerly as on the one that wins. Automate the level, the trigger mechanics and the size. Keep the read, and automate everything that happens after you have made it.
INFO
Quick answer. Automate the mechanical layers of an SMC trade: alerting on a level, sizing from risk percent, placing and managing the order, and journalling the outcome. Keep the structural read discretionary. The practical shape is a TradingView alert that fires when price reaches a level you chose, a webhook that carries it to your account, and a journal that records why.
An order block is not a shape. It is the last opposing candle before a move that did something — took liquidity, broke a level that mattered, left an imbalance that has not been revisited. Every part of that definition contains a judgement about context, and the judgement is what separates a trader with a hit rate from an indicator with a signal count.
You can encode a decent approximation. Most SMC indicators, ours included, do exactly that: they find the candle, check the displacement that followed, measure the imbalance and mark the zone. That is useful. What it cannot do is know that this particular break came into a weekly level from the wrong side, or that the session is thin because half of Asia is on holiday, or that the release in forty minutes is the one that has moved this pair three times this quarter.
The honest framing: an indicator narrows your attention from a whole chart to four candidates. Choosing between the four candidates is the work, and it remains yours.
WARNING
The failure mode when people automate the read is not that the system loses. It is that the system takes forty trades a week instead of four, at a slightly worse average, and the compounding of costs and small errors does the rest. Trade count is the tell.
Alerting. The highest-value automation in discretionary SMC, and the one people skip. You have chosen a level. You do not need to watch it. An alert on the level — on the chart, on your phone, in a channel — converts hours of screen time into a notification. It also does something subtler: it forces you to commit to the level in advance, in writing, at a moment when you are calm.
Sizing. Position size from a risk percentage and the actual distance to invalidation is pure arithmetic, and humans are unreliable at arithmetic under time pressure. The important detail is which price the percentage is measured against. Our sizing measures risk percent on the actual fill, not on the intended entry, because on a fast level those are not the same number and the difference is entirely your risk. A 12-point slip on a 40-point stop is 30 per cent more risk than you authorised.
Order placement and mechanical management. Placing the entry, attaching the stop, attaching the target, and moving the stop according to a rule you wrote down beforehand. "Rule you wrote down beforehand" is doing all the work in that sentence. If the management rule is really "I move it when it feels right", automating it will produce a machine that does the wrong thing quickly rather than a machine that does the right thing.
Journalling. Entry, stop, target, close reason, and whether each of your own rules was followed on that trade. This is the layer with the highest ratio of value to effort, because it is the only one that tells you later whether the other three are working. Our journal syncs those fields from the account and grades adherence per rule, so the question "do I actually take the setups I said I would" has an answer rather than an impression.
Replication. Once a decision exists on one account, moving it to your other accounts is mechanical by definition. This is a Provider and Follower relationship — one account originates, the others receive — and any account can be either.
The concrete route for a TradingView-based SMC trader looks like this.
Four things about that path are worth knowing before you build on it.
The alert carries intent, not a position. A TradingView alert is a message. It says "this condition happened on this symbol at this price". What it becomes — market order, limit order, which account, what size — is decided on our side from settings you configured in advance, not from the chart.
Limit and stop entries are watched, not fired and forgotten. If your alert describes an entry at a level price has not reached yet, the order is held and watched until it fills or expires. It does not silently become a market order at whatever price happens to be trading.
Close alerts are first-class. An alert that says "close" closes the position on the linked accounts. An SMC strategy that has an exit condition needs this to be as reliable as the entry, and it is the part half-built webhook setups forget.
Every skip has a reason. If a trade is not taken — the risk check failed, the slippage limit was exceeded, the account was at its daily halt, the instrument was not mapped — it appears in the Cockpit as a held or skipped trade with the reason next to it. A system that silently does nothing is indistinguishable from a system that is broken.
Here is one arrangement, offered as an illustrative example rather than a recommendation, of what a discretionary SMC trader can hand over and what they keep.
The column that matters is the second one. Two rows are yours. Everything else is machinery, and machinery is better at machinery than you are at three in the afternoon on the fourth day of a losing week.
The reason to automate the mechanical layers is not speed. It is that the mechanical layers are exactly where discipline fails, and they fail in a specific direction.
Nobody oversizes when calm. They oversize after two losses, on the trade that will get it back. Nobody skips their stop on the setup they described in their plan; they skip it on the one they took because they were bored. A system that computes size from a fixed percentage and attaches the stop before you can see the position move does not stop you from being human. It stops your being human from reaching the account.
That is a modest claim and it is the true one. Automation does not create an edge. It stops the edge you already have from leaking out through the parts of trading that are just admin.
The mechanical layers can — alerting, sizing, order placement, partial exits, journalling and replication to other accounts. The structural read generally cannot, because an order block's validity depends on context an indicator does not hold: session liquidity, higher-timeframe location, and what is on the calendar. Fully automated SMC systems exist, but they are different strategies from discretionary SMC rather than the same strategy running by itself.
Point the alert at a webhook. You link the account once, configure size, risk limits and slippage tolerance on our side, then include the webhook URL in the alert's settings. When the condition fires, TradingView posts the message, the relay validates and sizes it, and the order goes to each linked account. Entries at a price not yet reached are held and watched until they fill or expire.
No for the common case. TradingView's alert dialog can fire on a drawing, a price level or a built-in condition without any code. Pine becomes necessary when the condition is something only your own logic can express — a break of structure confirmed by displacement, for instance — and at that point you are writing an indicator rather than an automation.
Anything whose rule you cannot state in a sentence before the trade exists. If you cannot write down what would make you exit early, do not automate an early exit, because what gets encoded will be a guess at your own behaviour rather than your behaviour. The same test applies to adding to positions and to trading a release.
Not in itself. Firms restrict specific behaviours — copying from another person's account, coordinated group trading, certain latency and news practices — rather than automation as a category, and many permit Expert Advisors and alert-driven execution outright. Terms vary and change; as of September 2026 the sensible step is to read your firm's current automation clause before you connect anything.
At the end, and it is the layer that makes the rest improvable. When entry, stop, target and close reason arrive from the account automatically, and each of your own rules is graded as followed or not, you can answer questions that opinion cannot settle: whether the setups you skip outperform the ones you take, and whether your rule-breaking trades are the losses or the wins.
If the boundary question is what you came for — where a trading system properly ends and a trading decision begins — /blog/what-should-never-be-automated argues that line in more depth and is worth reading before you build anything. For the structural side, the read this post deliberately leaves with you, /blog/smc-orderblock-explained covers what makes an order block worth trading in the first place.
