Beta Lab / Gold Vigil
Gold Vigil
XAUUSD fractal swing engine — enters a breath before the level breaks. Open to everyone.
Two things before it will run
- Allow it to reach the licence server. In MetaTrader:
Tools → Options → Expert Advisors, tick
“Allow WebRequest for listed URL” and add
https://mcpautomation.io. Without this MetaTrader blocks the check and the EA will not start. - Paste your beta key into
InpLicenseKeyin the EA's Inputs tab. Your key appears on this page the moment you are approved.
The key is tied to your account and 2 MT5 accounts. Sharing it switches it off.
What it does
Gold Vigil trades gold against the US dollar on MetaTrader 5. It watches the most recent
standing swing high and swing low on the hourly chart and sets a stop order just INSIDE
each level — a breath before the level gives way, rather than after it breaks. Every
trade carries a stop loss. No grid, no martingale, no averaging into losers.
How a trade happens:
• Every M15 bar it scans H1 for the newest swing high and swing low — a bar whose high
(or low) beats the five bars either side, and which nothing newer has taken out.
• It places a buy stop 0.08% BELOW the swing high, or a sell stop 0.08% ABOVE the swing
low, and moves that order onto a newer level as one forms.
• Stop loss 2% of price, take profit 0.7% of price. Every distance is a percentage, so
nothing needs re-tuning as gold re-prices from $1,800 to $4,300.
• Break-even at +0.1%. A trailing stop from +0.2% that stops trailing at +1% so a big
move is left to run. A "salvage" exit that walks the target back once a trade is 1%
underwater, to leave on a bounce instead of at the full stop.
• Unfilled pending orders expire after 15 hours.
What it did in testing — MT5 strategy tester, every tick based on REAL TICKS, XAUUSD on
RoboForex-Pro, $100,000, default settings, measured 21 October 2026:
last 3 months +1.1% PF 1.04 max equity drawdown 7.6% 334 trades
last 12 months +53.2% PF 1.45 max equity drawdown 7.7% 1,130 trades
all 4.8 years +170% PF 1.21 max equity drawdown 33.0% 6,266 trades
Read all three lines, not just the middle one. The last twelve months are the best
stretch this engine has ever had. But 2022 and 2024 were flat for the entire year, the
worst drawdown was a third of the account, and the last three months have been
break-even. A back test is run with hindsight on data that already happened; live is
always worse. How much worse is what this beta is for.
One more thing worth understanding before you judge it. Gold Vigil wins about 87% of its
trades, which sounds extraordinary until you notice it risks 2% to make 0.7%. That ratio
needs roughly a 74% win rate just to break even. The real margin is thirteen points, not
eighty-seven, and one full stop loss erases about three winners.
THE VIGIL TRIALS — EARN YOUR RANK
Testing this earns XP, XP earns rank, and rank earns things that are not for sale: early
builds, your settings shipped inside the EA as a named preset, your name credited in the
EA itself, and a free lifetime licence for Gold Vigil v1.0 if it is ever released
commercially. Fourteen quests, six ranks, five Keeper seats ever.
The quest board https://mcpautomation.io/beta/gold-vigil/quests
The first 25 approved testers keep the Founding Tester mark permanently, whatever rank
they finish on.
THE FULL MANUAL AND THE LICENCE
Everything above in detail — every setting, how to install it, how to make it calmer, and
a troubleshooting table — is in the manual. Read it before you attach the EA.
Manual https://mcpautomation.io/assets/goldvigil/GoldVigil_Beta_Manual.pdf
Licence https://mcpautomation.io/assets/goldvigil/GoldVigil_Beta_Licence.pdf
The licence is the Gold Vigil Beta Software Licence Agreement v1.0, which you accept by
running the software. In short: free, personal, revocable, for testing only; do not share
the file or your key; unfinished software with no warranty; each licence check sends your
key, account number, broker, account type, currency, leverage, terminal build and a
machine identifier; and if you run it on real money, any loss is yours.
standing swing high and swing low on the hourly chart and sets a stop order just INSIDE
each level — a breath before the level gives way, rather than after it breaks. Every
trade carries a stop loss. No grid, no martingale, no averaging into losers.
How a trade happens:
• Every M15 bar it scans H1 for the newest swing high and swing low — a bar whose high
(or low) beats the five bars either side, and which nothing newer has taken out.
• It places a buy stop 0.08% BELOW the swing high, or a sell stop 0.08% ABOVE the swing
low, and moves that order onto a newer level as one forms.
• Stop loss 2% of price, take profit 0.7% of price. Every distance is a percentage, so
nothing needs re-tuning as gold re-prices from $1,800 to $4,300.
• Break-even at +0.1%. A trailing stop from +0.2% that stops trailing at +1% so a big
move is left to run. A "salvage" exit that walks the target back once a trade is 1%
underwater, to leave on a bounce instead of at the full stop.
• Unfilled pending orders expire after 15 hours.
What it did in testing — MT5 strategy tester, every tick based on REAL TICKS, XAUUSD on
RoboForex-Pro, $100,000, default settings, measured 21 October 2026:
last 3 months +1.1% PF 1.04 max equity drawdown 7.6% 334 trades
last 12 months +53.2% PF 1.45 max equity drawdown 7.7% 1,130 trades
all 4.8 years +170% PF 1.21 max equity drawdown 33.0% 6,266 trades
Read all three lines, not just the middle one. The last twelve months are the best
stretch this engine has ever had. But 2022 and 2024 were flat for the entire year, the
worst drawdown was a third of the account, and the last three months have been
break-even. A back test is run with hindsight on data that already happened; live is
always worse. How much worse is what this beta is for.
One more thing worth understanding before you judge it. Gold Vigil wins about 87% of its
trades, which sounds extraordinary until you notice it risks 2% to make 0.7%. That ratio
needs roughly a 74% win rate just to break even. The real margin is thirteen points, not
eighty-seven, and one full stop loss erases about three winners.
THE VIGIL TRIALS — EARN YOUR RANK
Testing this earns XP, XP earns rank, and rank earns things that are not for sale: early
builds, your settings shipped inside the EA as a named preset, your name credited in the
EA itself, and a free lifetime licence for Gold Vigil v1.0 if it is ever released
commercially. Fourteen quests, six ranks, five Keeper seats ever.
The quest board https://mcpautomation.io/beta/gold-vigil/quests
The first 25 approved testers keep the Founding Tester mark permanently, whatever rank
they finish on.
THE FULL MANUAL AND THE LICENCE
Everything above in detail — every setting, how to install it, how to make it calmer, and
a troubleshooting table — is in the manual. Read it before you attach the EA.
Manual https://mcpautomation.io/assets/goldvigil/GoldVigil_Beta_Manual.pdf
Licence https://mcpautomation.io/assets/goldvigil/GoldVigil_Beta_Licence.pdf
The licence is the Gold Vigil Beta Software Licence Agreement v1.0, which you accept by
running the software. In short: free, personal, revocable, for testing only; do not share
the file or your key; unfinished software with no warranty; each licence check sends your
key, account number, broker, account type, currency, leverage, terminal build and a
machine identifier; and if you run it on real money, any loss is yours.
What we need you to find out
A tester told exactly what to look for reports something useful. A tester told nothing
reports whether it made money for two weeks, which tells neither of us anything.
Roughly in order of how much I need it:
1. DOES IT BEHAVE THE SAME ON YOUR BROKER? Every number above comes from one feed
(RoboForex-Pro, GMT+3). Spread, swap and symbol naming all differ elsewhere. Tell me
your broker, your symbol name, and whether you are seeing anywhere near 100-130
positions a month.
2. DOES IT START AT ALL? Brokers call gold XAUUSD, XAUUSD.m, XAUUSD.pro, GOLD and worse.
If it attaches and then never places a single order, that is the single most useful
thing you can tell me.
3. ARE THE STOP ORDERS BEING REJECTED? Some brokers enforce a minimum distance between
the current price and a pending order, and mine is unusually permissive. Anything in
the Journal tab about "invalid stops" or "invalid price" matters.
4. IS THE LOT SIZE SANE ON YOUR ACCOUNT SIZE? Risk is 1.65% of balance per trade by
default, and up to twenty positions can be open at once. On a small account the 0.01
minimum lot can force the real risk well above what you set.
5. DOES THE SALVAGE EXIT ACTUALLY FIRE? It is the least-tested mechanic in the EA. I want
to see trades that closed at a PARTIAL loss rather than the full 2% stop.
6. ANYTHING IN THE EXPERTS OR JOURNAL TAB THAT READS LIKE AN ERROR. Copy the line.
What I am NOT asking is whether it made money over two weeks. It earns in bursts and then
does nothing for months — that is visible in the yearly numbers above. A losing fortnight
is not a bug.
reports whether it made money for two weeks, which tells neither of us anything.
Roughly in order of how much I need it:
1. DOES IT BEHAVE THE SAME ON YOUR BROKER? Every number above comes from one feed
(RoboForex-Pro, GMT+3). Spread, swap and symbol naming all differ elsewhere. Tell me
your broker, your symbol name, and whether you are seeing anywhere near 100-130
positions a month.
2. DOES IT START AT ALL? Brokers call gold XAUUSD, XAUUSD.m, XAUUSD.pro, GOLD and worse.
If it attaches and then never places a single order, that is the single most useful
thing you can tell me.
3. ARE THE STOP ORDERS BEING REJECTED? Some brokers enforce a minimum distance between
the current price and a pending order, and mine is unusually permissive. Anything in
the Journal tab about "invalid stops" or "invalid price" matters.
4. IS THE LOT SIZE SANE ON YOUR ACCOUNT SIZE? Risk is 1.65% of balance per trade by
default, and up to twenty positions can be open at once. On a small account the 0.01
minimum lot can force the real risk well above what you set.
5. DOES THE SALVAGE EXIT ACTUALLY FIRE? It is the least-tested mechanic in the EA. I want
to see trades that closed at a PARTIAL loss rather than the full 2% stop.
6. ANYTHING IN THE EXPERTS OR JOURNAL TAB THAT READS LIKE AN ERROR. Copy the line.
What I am NOT asking is whether it made money over two weeks. It earns in bursts and then
does nothing for months — that is visible in the yearly numbers above. A losing fortnight
is not a bug.
What you need
• MetaTrader 5. Anyone can join — you do not need to own anything of mine first.
• An XAUUSD chart. Any timeframe: the EA uses its own H1, M15 and M1 internally and
ignores the chart period.
• DEMO is strongly recommended. The build does NOT block a real-money account — it
prints a five-line warning and raises an Alert, then runs. If you want it to refuse
outright, set InpDemoOnly = true. On a live account, any loss is yours.
• MetaTrader must be allowed to reach the licence server: Tools > Options > Expert
Advisors, tick "Allow WebRequest for listed URL" and add https://mcpautomation.io
• A terminal that stays on, so a VPS helps but is not required. It trades around the
clock apart from 30 minutes either side of the daily rollover.
• An account you are willing to let run for at least two weeks. It takes roughly 100-130
trades a month, so a few days tells us nothing.
• An XAUUSD chart. Any timeframe: the EA uses its own H1, M15 and M1 internally and
ignores the chart period.
• DEMO is strongly recommended. The build does NOT block a real-money account — it
prints a five-line warning and raises an Alert, then runs. If you want it to refuse
outright, set InpDemoOnly = true. On a live account, any loss is yours.
• MetaTrader must be allowed to reach the licence server: Tools > Options > Expert
Advisors, tick "Allow WebRequest for listed URL" and add https://mcpautomation.io
• A terminal that stays on, so a VPS helps but is not required. It trades around the
clock apart from 30 minutes either side of the daily rollover.
• An account you are willing to let run for at least two weeks. It takes roughly 100-130
trades a month, so a few days tells us nothing.
The rules, in plain words
- This is unfinished software. It has not been through a full release test and it may contain mistakes that cost money.
- Run it on a demo account. If you choose to run it on real money anyway, you do that entirely at your own risk and you accept every loss it causes.
- MCP Automation gives no warranty of any kind on a beta build and is not liable for any loss, missed trade or account breach that comes from running it.
- The build, its key and its settings are yours to test, not to share, resell, decompile or upload anywhere.
- Your key is tied to your account and stops working when the beta ends or if you are removed from the programme.
- You agree to post at least one honest test report. That is the whole price of admission.
- Nothing here is financial advice and no result, past or simulated, promises anything about the future.
Builds
1.01b
Current
GOLD VIGIL 1.00b — first beta build
=====================================
This is the first public build. There is no previous version to compare
against, so this changelog is a statement of what is in it.
THE ENGINE
• Finds the most recent *standing* swing high and swing low on H1 — a bar
whose high (low) beats the 5 bars either side of it, and which nothing
newer has taken out.
• Rescans every M15 bar and repositions its orders onto the newest level.
• Places a buy stop 0.08% BELOW the swing high, or a sell stop 0.08% ABOVE
the swing low. The offset is negative on purpose — it enters a breath
before the level gives way, not after.
• Stop loss 2% of price. Take profit 0.7% of price. Every distance is a
percentage, so nothing needs re-tuning as gold re-prices.
• Break-even at +0.1%, locking +0.025%.
• Trailing stop from +0.2%, trailing 0.2% behind price, and it stops
trailing at +1% so a big move is left to run.
• Salvage exit: once a trade is 1% underwater the target is walked down to
0.15% above price, to leave on a bounce instead of at the full stop.
• Unfilled pending orders expire after 15 hours.
• No new orders within 30 minutes either side of the daily rollover.
WHAT IS IN THIS BUILD THAT IS NOT IN THE ENGINE
• InpLicenseKey — your key from mcpautomation.io/portal/beta
• InpDemoOnly — true makes the EA refuse to start on a real-money account
• A chart panel showing build, licence state and account type
• A real-money warning: five lines in the Experts log plus an Alert. It
warns, it does not block. Use InpDemoOnly if you want it to block.
KNOWN BEHAVIOUR — NOT BUGS
• It trades a lot. Roughly 100–130 positions a month on XAUUSD.
• The win rate is high (86–90%) because it risks 2% to make 0.7%. That ratio
needs about 74% wins just to break even, so the real margin is much
thinner than the win rate suggests.
• It earns in bursts and then does nothing for months. In testing, 2022 and
2024 were both flat-to-negative for the whole year.
• Several positions can be open at once — up to 10 per side.
MEASURED BEFORE RELEASE
MT5 strategy tester, every tick based on real ticks, XAUUSD on RoboForex-Pro,
$100,000, default settings, 1 Jan 2022 → 21 Oct 2026:
all 4.8 years +170% PF 1.21 max equity drawdown 33.0% 6,266 trades
last 12 months +53.2% PF 1.45 max equity drawdown 7.7% 1,130 trades
last 3 months +1.1% PF 1.04 max equity drawdown 7.6% 334 trades
Read those three lines together, not separately. The last year was the best
stretch it has ever had. The last three months were flat. The 33% drawdown
is real and happened in 2022–2024.
A back test is not a promise. That is the whole reason this beta exists.
=====================================
This is the first public build. There is no previous version to compare
against, so this changelog is a statement of what is in it.
THE ENGINE
• Finds the most recent *standing* swing high and swing low on H1 — a bar
whose high (low) beats the 5 bars either side of it, and which nothing
newer has taken out.
• Rescans every M15 bar and repositions its orders onto the newest level.
• Places a buy stop 0.08% BELOW the swing high, or a sell stop 0.08% ABOVE
the swing low. The offset is negative on purpose — it enters a breath
before the level gives way, not after.
• Stop loss 2% of price. Take profit 0.7% of price. Every distance is a
percentage, so nothing needs re-tuning as gold re-prices.
• Break-even at +0.1%, locking +0.025%.
• Trailing stop from +0.2%, trailing 0.2% behind price, and it stops
trailing at +1% so a big move is left to run.
• Salvage exit: once a trade is 1% underwater the target is walked down to
0.15% above price, to leave on a bounce instead of at the full stop.
• Unfilled pending orders expire after 15 hours.
• No new orders within 30 minutes either side of the daily rollover.
WHAT IS IN THIS BUILD THAT IS NOT IN THE ENGINE
• InpLicenseKey — your key from mcpautomation.io/portal/beta
• InpDemoOnly — true makes the EA refuse to start on a real-money account
• A chart panel showing build, licence state and account type
• A real-money warning: five lines in the Experts log plus an Alert. It
warns, it does not block. Use InpDemoOnly if you want it to block.
KNOWN BEHAVIOUR — NOT BUGS
• It trades a lot. Roughly 100–130 positions a month on XAUUSD.
• The win rate is high (86–90%) because it risks 2% to make 0.7%. That ratio
needs about 74% wins just to break even, so the real margin is much
thinner than the win rate suggests.
• It earns in bursts and then does nothing for months. In testing, 2022 and
2024 were both flat-to-negative for the whole year.
• Several positions can be open at once — up to 10 per side.
MEASURED BEFORE RELEASE
MT5 strategy tester, every tick based on real ticks, XAUUSD on RoboForex-Pro,
$100,000, default settings, 1 Jan 2022 → 21 Oct 2026:
all 4.8 years +170% PF 1.21 max equity drawdown 33.0% 6,266 trades
last 12 months +53.2% PF 1.45 max equity drawdown 7.7% 1,130 trades
last 3 months +1.1% PF 1.04 max equity drawdown 7.6% 334 trades
Read those three lines together, not separately. The last year was the best
stretch it has ever had. The last three months were flat. The 33% drawdown
is real and happened in 2022–2024.
A back test is not a promise. That is the whole reason this beta exists.