v2.16.0

One bot per market in each account: a second bot on a market an account already trades is not let live

No strategy changes: there is no new strategy version.

Why

A tester ran an MNQ 5 min bot and an Autonomous MNQ 1 min bot on one paper account. The 5 min bot's card said $0 and 0 trades while its trade had been closed by something else, about $1,450 of the account's day belonged to no bot, and TradersPost rejected signals on both bots. The logs could not say who closed the trade, so two bots on one market was a strong suspect, not a proof. What is certain is how brokers work: your broker holds one position per market in each account, so one bot's exit closes the other's trade, and what each bot shows and what the account made then stop matching. 2.15.0 said so on the bot's card and in Go Live, and both bots kept running. So it is now a rule.

What changed

• Go Live does not put a second bot on a market an account already trades. In the Go Live sheet an account that another bot is already trading this market on is greyed, with who is on it written on its row ("MNQ · 5 min is already trading MNQ here"). One line under the list says why, and only appears when an account is taken. Go Live itself refuses it too, with nothing touched, for anything the sheet could not see.
• A new Autonomous job is not set up beside such a bot. The job check says which bot and what to do. A job you are editing is only warned, so a job whose bot already runs beside another stays editable.
• A different market on the same account is fine. MNQ and MES together work as before: they are different positions at the broker.
• So are two bots on one market whose hours never overlap, such as an Asia bot and a New York bot: each is flat when the other trades. A Europe bot and a New York bot overlap from 9:30 am to 10:55 am, so that pair is refused, and the refusal says the hours overlap. A bot with no hours, or one that lets a trade run past the end, can be holding at any hour, so it counts against every other.
• Bots that are already running are never touched. They keep the line on their card. A job that cannot put its bot live because of one stops and says why.
• It only knows this app's bots. Another computer, or a trade placed by hand in TradersPost, is invisible to it.

The words are short on purpose: what is taken, the one reason (your broker holds one position per market in each account, so one bot's exit closes the other's trade), what to do, and what is still fine.

Also: the note under a Heikin Ashi test

Test on Chart on a Heikin Ashi result was reported as "it didn't select it". It had: the strategy's Signal candles read Heikin Ashi (its settings and its line on the chart say so, and the app reads it back before it tests anything), and the test's figures equalled the search's to the dollar. What the chart shows is ordinary bars, on purpose: a chart in Heikin Ashi style makes TradingView fill orders at prices nobody can trade. The Test & adjust sheet of a Heikin Ashi result now says where Heikin Ashi is set and why the bars stay ordinary. A Bars result's sheet is unchanged.

What did not change

The search, the results and their order, the engine and the job's rules, saved bots and jobs, the data that is sent, and how a bot goes live once it is allowed to.

Checked

• The automated tests: all pass (4,396 run, 7 skipped by design, as before). The real Go Live code was run over stand-ins: refused beside another bot on the same market and account with nothing touched (the bot stays as it was, no deploy begins, Chrome is not asked for anything), allowed on another account, on another market, beside a bot that is off, and by hours that never overlap; a bot is never refused by its own history (Try Again).
• 28 rules of the new code were cut out of it, one at a time, and a test failed every time (how hours are read and compared, which bots count, what each message says, what the Go Live sheet greys and unticks).
• On the real app, restarted onto this build: Go Live on a second Micro Gold bot, beside the Micro Gold job's bot on the same paper account, was refused with the words above; the bot stayed as it was, no deploy began, and it was removed afterwards. The Go Live sheet for it showed that account greyed with "MGC · 5 min · Whole day is already trading MGC here", the one reason under the list, and the other accounts free.
• The Heikin Ashi note on a real Heikin Ashi result's sheet, and no note on a Bars result's.

Not run

• A Go Live that is allowed, on the real app (another account, another market, hours that never overlap). They are covered by the tests over stand-ins, not done with a real alert.
• A job's swap stopped by such a bot on a live account. The words come from the same check; the swap itself was not run into it.
• Windows and Linux. Run on macOS only.

---

Unsigned build.

Windows first install:

1. Run AutoBotDeploy-Setup-2.16.0.exe.
2. SmartScreen will warn because the build is unsigned — More info → Run anyway.

macOS first install:

1. Pick the right DMG — Apple Silicon = *-arm64.dmg, Intel = the one without -arm64.
2. Drag AutoBotDeploy into Applications (do not run from the DMG).
3. Open it once. When macOS says it cannot be opened: System Settings → Privacy & Security → scroll down → Open Anyway, then open it again.
4. If it still refuses: Terminal → xattr -cr /Applications/AutoBotDeploy.app then open it again.

Updates download and verify automatically; install them yourself when flat.

v2.15.2

A new job goes with the market again: the engine is a choice, not the default

No strategy changes: there is no new strategy version.

What went wrong

2.14.0 added an engine that judges whether a result is a long-run edge, and made it the default of a new Autonomous job: rank results by it, and leave out any it is not 80% sure of. Autonomous mode is for riding what is working now (search the last few days, put the best result on, search again after losses), and on a few days of history nothing passes an 80% bar: a result with a measured loss needs about 14 trades, a run with no losing trade about 44. A new job could wait for ever. A Micro S&P job held on a +$1,451 result (10 trades, no losing trade) because the engine was 58% sure.

What changed

• A new job ranks by the app's score and asks for no confidence, as jobs did before 2.14.0.
• The engine is two choices in the job form: the rank "Expected profit after a cut for few trades (a long-run view)" and the switch "Be sure it makes money", suggested at 80 and now said to be strict. Neither is on for a new job.
• The engine is off the screens a person reads results on. The line on a result's card in Find a Bot, the Engine column of the cycle table and the engine's clause in the scorecard are gone. A decision says the engine's words only for a job that asked for the engine.
• Jobs you already made keep what they were saved with. If a job you made on 2.14.0 to 2.15.1 is holding because of the 80% bar, open it, switch Be sure it makes money off (in Which result counts) and set Rank by to The app's score.
• Each pick still keeps the engine's prediction, which AutoBot's usage data (private, on by default, a switch in Settings) sets beside what the pick then made, so it can be checked. It is not shown on a result card.

What did not change

The search, the order of results on the Find a Bot page, the score, the stability badge (per trade), the rule that a result's last trades must have won, the 2.15.x data that is sent, and every saved bot and job.

Checked

• The automated tests: all pass (4,381 run, 7 skipped by design, as before). The result that raised this is a test: a Micro S&P result of +$1,451 on 10 trades with no losing trade is taken by a new job and held by a job that asks the engine to be 80% sure, with the engine's own reason.
• On the real app: the dev app restarted onto this build and the two Autonomous jobs on it were switched back to the score with no bar. The Micro S&P job searched again (Run 37), took Width 8 · Period 5 · Target 0.3R · Stop 0.6% · Trend 100 (the result it had been holding on) and put its bot live on the paper account.

Not run

• A job created fresh under the new defaults. The Micro S&P job was switched to the same rules, which is what a new job now starts with.
• Windows and Linux. Run on macOS only.

---

Unsigned build.

Windows first install:

1. Run AutoBotDeploy-Setup-2.15.2.exe.
2. SmartScreen will warn because the build is unsigned — More info → Run anyway.

macOS first install:

1. Pick the right DMG — Apple Silicon = *-arm64.dmg, Intel = the one without -arm64.
2. Drag AutoBotDeploy into Applications (do not run from the DMG).
3. Open it once. When macOS says it cannot be opened: System Settings → Privacy & Security → scroll down → Open Anyway, then open it again.
4. If it still refuses: Terminal → xattr -cr /Applications/AutoBotDeploy.app then open it again.

Updates download and verify automatically; install them yourself when flat.

v2.15.1

The line under a result now says what the test showed, beside what it is worth

No strategy changes: there is no new strategy version.

What changed

The engine line under each result (added in 2.14.0) said "Worth about +$17 a trade, about +$30 a day after the cut · 75% sure it makes money". Next to a net profit of +$1,423 that read as a contradiction, because the figure it had been cut from was not on the card. It now says it:

Test showed +$158 a trade; allowing for luck, worth about +$17 a trade (about +$30 a day) · 75% sure it makes money

Only the words changed: the numbers are the same, and the tooltip is the same. Read it as before: the test showed +$158 a trade, but it is the best of many tries on the same days and rests on 9 trades, so the engine keeps about 10% of that, which is about +$17 a trade. "75% sure it makes money" means 75% sure the real edge is above zero. It is not a 75% win rate.

What did not change

The engine's arithmetic, the ranking of results, what an Autonomous job asks for (a new job still asks to be 80% sure, so a result that rests on about 9 trades will not pass it), the cycle table and every saved setting.

Checked

• The automated tests: all pass (4,379 run, 7 skipped by design, as before). The new words are held to the real engine's figures for the card that raised this (Micro S&P 500, 5 min, +$1,423 on 9 trades), for a result whose test length is not known (no empty brackets) and for a result that lost money in its test.
• On the real app: a real result on the Find a Bot page reads "Test showed +$145 a trade; allowing for luck, worth about +$6 a trade (about +$11 a day) · 58% sure it makes money", and wraps to three lines on the card without clipping.

Not run

• Windows and Linux. Run on macOS only.

---

Unsigned build.

Windows first install:

1. Run AutoBotDeploy-Setup-2.15.1.exe.
2. SmartScreen will warn because the build is unsigned — More info → Run anyway.

macOS first install:

1. Pick the right DMG — Apple Silicon = *-arm64.dmg, Intel = the one without -arm64.
2. Drag AutoBotDeploy into Applications (do not run from the DMG).
3. Open it once. When macOS says it cannot be opened: System Settings → Privacy & Security → scroll down → Open Anyway, then open it again.
4. If it still refuses: Terminal → xattr -cr /Applications/AutoBotDeploy.app then open it again.

Updates download and verify automatically; install them yourself when flat.

v2.15.0

Help improve AutoBot sends more, so the selection engine can be checked against what really happens; and two things a tester's videos showed

No strategy changes: there is no new strategy version.

Help improve AutoBot: what is new in what it sends

It is still on by default, private (only the people who make AutoBot see it) and switched off in Settings > Help improve AutoBot, and the one line there says what is sent. The line has changed, because more is sent now:

AutoBot sends the settings of your bots and Autonomous jobs, the test figures they were chosen on and the best results each search found, what your bots have made so far (totals), when a bot or job changes state or runs into trouble (a daily limit stopping trading included), when a search fails, and the app version, under a random ID or a name you choose, to help improve it. Never your computer’s name, account names, logins or individual trades. Like any website, the database can see your internet address.

Settings > Help improve AutoBot > What is sent lists every item. What is new:

• The app version is on everything. It was on the install and on every event; it is now on each bot, job, cycle and search too, because what a number means depends on the version of the logic that made it.
• What a bot was started on, beside what the engine expected of it. The test figures its settings were chosen on (profit, profit factor, win rate, trades, drawdown, how long the test was, and for an Autonomous bot the app's grade and score) and the engine's prediction as it was when the bot was started, so the prediction can be checked against what the bot then made. They are fixed at the bot's first row: a later look never rewrites them.
• A little more about what a bot has made since it went live. The wins and the losses added up, the deepest dip and the longest losing run. Still totals, sent when a go-live starts, when it ends and about every three hours between: never a trade.
• Each Autonomous cycle that ends. Why it ran, what it decided and why (one fixed word), how many results it looked at, the best five (their settings and test figures, whether each qualified and the fixed reason if not), what it chose, and what the settings it replaced went on to make (trades, wins, profit): the prediction and the result of a pick in one row.
• Each search that finishes. What was asked, how many settings it tried, how long it took, a summary of how the results of each chart did (how many made money in their test, the middle and the spread of profit a trade, how many were stable) and its best three. These are tests of settings over market history, not your trades.

What it is for. The engine added in 2.14.0 says what a result is worth going forward. Whether it is right can only be known by setting what it expected beside what picks then made, and a pick from one install is too few to tell. This is how that gets measured, and how it gets corrected if it is wrong. It cannot create an edge, and a few weeks of a handful of installs can only say how much of a test survives overall, not by market.

What did not change. Never your computer's name, account names, logins, individual trades or any text AutoBot writes (the reasons above are matched to fixed words and never sent as sentences). Names of bots and jobs are never read. Updating does not send your past cycles and searches: only those that end after the update are sent (a bot that is live and a job that exists are sent as they are, as before, now with the new fields). Nothing pops up and nothing asks. Delete my data removes the new rows with the rest, and switching it off sends nothing more.

Two things a tester's videos showed

• A trade closed outside a bot is no longer shown as "$0, 0 trades". A bot's trade was closed by something else (by hand, a prop firm's flatten, or another bot's exit on the same market), and its card said Today $0, Since live $0, 0 trades while the account had taken a loss of about $1,450 that no bot owned. It is now counted as a trade with an unknown result, on the day it was entered: that day and the total read as a dash with the reason, and the line says where the money went (the account's figure). A trade seen flat only once stays out of every figure, as before.
• Two bots on one market in one account are now said. A broker holds one position per market per account, so two such bots close each other's trades. AutoBot knows every bot's market and accounts, so it says so: on a bot's card while both are on the air, in Go Live when a chosen account already has another live bot on that market, and when a job is set up. It is a warning, never a refusal.

What did not change

Bots, alerts, limits, results, saved settings and Autonomous jobs carry over. The search, the engine and how a job chooses are exactly as in 2.14.0.

Checked

• The automated tests: all pass (4,379 run, 7 skipped by design, as before). About 230 times a rule of the new code was cut out of the built program (each row's fields, the fixed words, the checks, the snapshot, the analysis) and a test failed every time except four times where cutting changes nothing that can be sent.
• On the live database, after the tables were added: the contract check posts one row of every kind from the app's own code under a throwaway install, reads each table back, and passed in full: every table took its rows, the new columns came back with what was sent, no sentence or name is anywhere in what the database holds, no other install and no missing key can read any table, the new summary views refuse the public key, a bad row is refused, and a delete removes every table's rows.
• On the real app: the app ran on a stand-in database (the notice in Settings and About, the What is sent sheet with its five items, the install row saying notice 2 and the app version), and Brent's own running install, restarted onto this code, sent its live MGC bot's row with the app version, the figures it was chosen on and the engine's prediction (it keeps 30%, about +$43 a trade) and its job's rule. His real saved searches and cycles were replayed through the new row builders and the database's checks: all pass, well inside the size limits.
• The two tester items: tests for the figures and words of the card, for the same-market warnings on the card, in Go Live and when a job is set up, and for the world these read from.

Not run

• A real cycle and a real search ending after the update and arriving as rows. They come when a job's next cycle ends. The rows are covered by tests and by the replay of real saved ones above.
• The reading of the data. It needs weeks of picks (about 30 pooled before the overall answer is better than a guess). The first report is scripts/usage-analyse.mjs, which says up front when there is too little to conclude.
• The two tester items on a live account. The closed-outside figures and the same-market warnings were not seen on the real app: that needs a trade closed by something else, and two bots on one market in one account.
• Windows and Linux. Run on macOS only.

---

Unsigned build.

Windows first install:

1. Run AutoBotDeploy-Setup-2.15.0.exe.
2. SmartScreen will warn because the build is unsigned — More info → Run anyway.

macOS first install:

1. Pick the right DMG — Apple Silicon = *-arm64.dmg, Intel = the one without -arm64.
2. Drag AutoBotDeploy into Applications (do not run from the DMG).
3. Open it once. When macOS says it cannot be opened: System Settings → Privacy & Security → scroll down → Open Anyway, then open it again.
4. If it still refuses: Terminal → xattr -cr /Applications/AutoBotDeploy.app then open it again.

Updates download and verify automatically; install them yourself when flat.

v2.14.0

The selection engine: what a result is worth going forward, in Find a Bot and in Autonomous

No strategy changes: there is no new strategy version.

What it is

A search tries 81 or more settings on the same few weeks and shows the best. The best of many always looks good, so a result shows more than it is worth, and the fewer trades it stands on, the more that is: a profit factor of 7.98 on 8 trades is not the same thing as a profit factor of 5 on 41. AutoBot's selection engine works out, for every result, what it is worth going forward. It runs on your computer from the numbers TradingView's Strategy Tester already gave (no cloud AI, nothing new is sent anywhere). It takes the average trade the test showed (in units of the result's own average losing trade, so a tight stop and a wide one compare), works out how noisy a trade is from the sizes of its wins and losses, and cuts the edge by how few trades it stands on. It then says how sure it is that what is left is above zero, and what that is worth in dollars a trade and a day.

On the result that started this (Micro Gold, 5 min, +$5,990, profit factor 5.15, 41 trades): it keeps about 30% of what the test showed, which is about +$43 a trade and +$83 a day, and is better than 99% sure the edge is above zero. A no-loss run on 8 or 12 trades, or a profit factor of 7.98 on 8, is not that sure.

In Find a Bot

• Every result card has a line under its net profit: "Worth about +$43 a trade, about +$83 a day after the cut · 100% sure it makes money", with a tooltip that says what was cut and why.
• The search and the order of the results are unchanged. The list is still ranked by grade, then score. The engine adds a line; it does not re-rank the page.
• The Stable / Mixed / Lone spike badge now compares per trade. It used to compare the neighbours' total profit and count a neighbour under 5 trades as not having made money, so it read how often a setting trades. The result above was "Mixed" because its five nearby settings had traded 3 to 9 times for +$582 to +$1,773, though all five made money (per trade $65 to $305, against $146 for the result). A neighbour with 1 or 2 trades is now left out as no evidence; a setting that never traded still counts as not having made money. That result now reads Stable.

In Autonomous

• A new job ranks by the engine and asks to be sure. Rank by has a new first choice, Expected profit after the engine's haircut (the default for a new job): it puts the result with the most evidence first, not the highest score. Be sure it makes money is a switch and a box (a new job asks for 80%): a result the engine is less sure of than that is left out, with its reason in words.
• A job you already have is unchanged. It keeps its own ranking and asks nothing extra (it has no such field in its file, and reads as off). The one thing that reaches it is the per-trade badge above, which matters if it has Stable only on: more results read Stable.
• The cycle table has an Engine column, and when a job chooses it says what the engine expects ("The engine expects about +$43 a trade (it keeps 30% of the +$146 the test showed) and is 100% sure it makes money.").
• The forward log. Each pick keeps the engine's prediction. When its settings are replaced, the cycle already records what they made live; the scorecard on the Autonomous page now sets the two side by side ("The engine had expected +$43 a trade of them"). After about 30 judged picks the engine re-fits the one constant its haircut rests on, pulled back toward the starting number by how few picks there are; until then it uses the starting numbers and claims nothing more. This is also how we will find out whether picking beats a random pick at all.

What it cannot do

• It cannot create an edge. The lab found this strategy at about +0.05 to +0.08R a trade when it works and about zero in the current regime. A month in which nearly every setting makes money (as the last one did on Micro Gold) is what pays, not the pick.
• The dollars are before commission and slippage, like everything the search shows. Costs are not charged yet.
• It judges from the summary numbers. It cannot see the shape of the equity curve (profit from three lucky trades, or only the first half making money). Reading every trade of the top few results is the next build.

What did not change

Bots, alerts, limits, results and saved settings carry over. Jobs keep their rules as saved. Help improve AutoBot and the community sharing are exactly as before.

Checked

• The automated tests: all pass (4,299 run, 7 skipped by design, as before). Each part of the engine and each rule it added was cut out of the code in turn (24 cut-outs) and a test failed every time; the ones that did not at first (the noise floor, the cautious figure, the estimate on a result) got tests until they did.
• The numbers are the real ones. The engine was run over all 81 settings of the Micro Gold search it was made for: only the 41-trade result and one 11-trade result clear 80%; the 8-trade no-loss runs and the profit factor of 7.98 on 8 trades do not; the ranking puts the 41-trade result first, then a 52-trade profit factor of 1.37.
• A real bug found before release: the prediction on a pick was dropped when the app closed, which would have left the forward log logging nothing. It is kept now, whole or not at all, and tested through the real file.
• On the real app: the line and the badge on the saved search it was made for (Run 34), the new switch and box and the rank list in the job's edit form, and the job that started this switched to the engine.

Not run

• A job choosing under the engine over a full cycle on a real account. The Micro Gold job is on it now; its next search is after its next losses. The choosing is covered by tests.
• The re-fit. It needs about 30 judged picks; there are not that many yet, so it is covered only by tests with made-up picks.
• Windows and Linux. Run on macOS only.

---

Unsigned build.

Windows first install:

1. Run AutoBotDeploy-Setup-2.14.0.exe.
2. SmartScreen will warn because the build is unsigned — More info → Run anyway.

macOS first install:

1. Pick the right DMG — Apple Silicon = *-arm64.dmg, Intel = the one without -arm64.
2. Drag AutoBotDeploy into Applications (do not run from the DMG).
3. Open it once. When macOS says it cannot be opened: System Settings → Privacy & Security → scroll down → Open Anyway, then open it again.
4. If it still refuses: Terminal → xattr -cr /Applications/AutoBotDeploy.app then open it again.

Updates download and verify automatically; install them yourself when flat.

v2.13.2

A way out of Pause all in Autonomous, and Autonomous takes a strong result with mixed neighbours

No strategy changes: there is no new strategy version.

Autonomous: new jobs skip only lone spikes

The rule Stable only in Which result counts is now three choices, under Its nearby settings:

• Skip lone spikes: the default for a new job. Only a result whose nearby settings made no money, or a tenth of what it did, is left out. A result with mixed neighbours is kept.
• Stable only: what every job had until now. Only a result whose nearby settings did well too.
• Do not look: the best result counts whatever the settings around it did.

A job you already have keeps what it had. One with Stable only on stays on Stable only, and one with it off stays on Do not look. Nothing is loosened behind your back; change it in Edit if you want the new default.

Why: a job's search listed a result with a profit factor of 5.15 over 41 trades (+$5,990, 88% wins, graded Strong, its last three trades won) and refused it for one rule, "its nearby settings are Mixed, not Stable". The job held with no bot. It was a paper job, and after it was switched to Skip lone spikes its next search chose that result and put it live.

Know this: the Stable / Mixed / Lone spike badge compares the neighbours' total profit with the result's, and counts a neighbour with fewer than 5 trades as not having made money. So a result that trades far more often than its neighbours can read "Mixed" while each neighbour made as much per trade. On that very result all five neighbours made money (per trade, $65 to $305, against $146 for the result), but they had traded 3 to 9 times against 41. The badge is not changed in this version: Skip lone spikes loosens the line but still uses it, so a result that trades far more often than its neighbours can still be left out. Comparing per trade, with a margin for how few trades a neighbour made, is next.

When nothing qualifies, the card says why the closest result was left out

Before: "Nothing qualified: 35 results listed, and the commonest reason was graded Weak, below Good (22 results)." That named the commonest reason across all of them, which was not why the best one was refused. Now it names the closest result, with its figures, and why it was left out ("The closest, Width 4 · Period 8 · Target 0.5R · Stop 0.6% · Trend 100 (score 98, +$5,990, profit factor 5.15, 41 trades), was left out because its nearby settings are "Mixed", not "Stable""), then the commonest reason across all of them. When it is the same reason, it is said once, with the count.

Pause all has a way out on the Autonomous page

If you press Pause all and then delete the bots it paused, Pause all stays on and holds every Autonomous job back. A job made after that said "Waiting · Pause all is on. It carries on by itself when that clears", and the only Resume all was on Home, so there was no way to end it from where the job was. Now:

• The job's card says Nothing starts until you press Resume all, and has a Resume all button. Only a job waiting on Pause all has one; a job waiting for Chrome or for an account stopped for the day still says it carries on by itself.
• The Pause all banner, with its Resume all, is on the Autonomous page and on an empty My Bots, as well as on Home. With no bot left it says "Pause all is on... No bot is live, but nothing starts while it is on, Autonomous jobs included. Resume all lifts it." instead of "All bots paused".
• The notice a job gives after ten minutes of waiting says the same.
• Resume all with no bot left lifts the pause at once and switches nothing on or off at TradersPost.

What did not change

Bots, alerts, limits, results and saved settings carry over. Jobs keep their rules as saved. Help improve AutoBot and the community sharing are exactly as in 2.13.1.

Checked

• The automated tests: 4,282 run, 4,275 pass, none fail (7 are skipped by design, as before). The new rules were cut out of the code in turn (28 cut-outs) and a test failed every time; one test that did not notice a missing choice on the form was tightened until it did.
• Real code for Resume all: with every bot deleted, Resume all lifts the pause, leaves nothing unfinished on disk and clicks nothing at TradersPost, with and without the deleted bots' records still listed (run over the app's own pause code, not a copy).
• On the real app: the Pause all banner and the job's Resume all button were looked at on screen on Autonomous, My Bots and Home with no bot left. The job whose result was refused was then switched to Skip lone spikes, its next search listed the same result, and the job put it live on the paper account.

Not run

• A swap under the new rule. The real job was making its first Go Live; replacing a live bot's settings under Skip lone spikes is covered by tests, not run on a real account.
• The new "closest result" sentence on a real search. It is covered by tests; it appears the next time a real search finds nothing.
• Windows and Linux. Run on macOS only.

---

Unsigned build.

Windows first install:

1. Run AutoBotDeploy-Setup-2.13.2.exe.
2. SmartScreen will warn because the build is unsigned — More info → Run anyway.

macOS first install:

1. Pick the right DMG — Apple Silicon = *-arm64.dmg, Intel = the one without -arm64.
2. Drag AutoBotDeploy into Applications (do not run from the DMG).
3. Open it once. When macOS says it cannot be opened: System Settings → Privacy & Security → scroll down → Open Anyway, then open it again.
4. If it still refuses: Terminal → xattr -cr /Applications/AutoBotDeploy.app then open it again.

Updates download and verify automatically; install them yourself when flat.

v2.13.1

Help improve AutoBot (new, on by default), the trading-hours blocks in Autonomous, and a hand Turn Off that holds

No strategy changes: there is no new strategy version.

Help improve AutoBot: new, and on until you switch it off

AutoBot sends the settings of your bots and Autonomous jobs, their result totals, when a bot or job changes state or runs into trouble (a daily limit stopping trading included), when a search fails, and the app version, under a random ID or a name you choose, to help improve it. Never your computer’s name, account names, logins or individual trades. Like any website, the database can see your internet address.

This is new in this version and it is on by default. It is private (only the people who make AutoBot see it) and separate from the community sharing of 2.13.0, which is unchanged and still off until you switch it on. Nothing pops up: it is a row in Settings > Help improve AutoBot, with the line above beside its switch, and the same line in About. The Home screen’s first-run card and the places that said “nothing is shared unless you switch sharing on” now say what is true.

• Switch it off there and nothing more is sent. It stays off after a restart and an update, and a settings file that is damaged or cannot be read counts as off, never as on. What was sent before stays until you press Delete my data, which removes it from the database (it needs no account: your computer holds the key) and switches sending off. If the database cannot be reached the delete is kept and tried until it works.
• Your ID is a random name like “Calm Heron 7F3A”, made on this computer the first time AutoBot starts, shown with the first characters of its key. It is what you tell support, so they can find what you see. Change name lets you type your own instead (3 to 32 letters, digits, spaces, dots, dashes and underscores), or roll another random one.
• What is sent, in full, is in the sheet What is sent beside the switch. For each time a bot goes live: its market, timeframe and trading hours, its five strategy settings (Width, Period, Target, Stop, Trend), the risk per trade and contract cap, whether it is manual, run by an Autonomous job or a shared bot, whether its account is paper or live, when that go-live began and ended, and how many trades it has closed, how many won and the profit so far: totals, never the trades themselves, sent when the go-live starts, when it ends and about every three hours between, so they cannot be read as a record of trades. For each Autonomous job: its market, timeframe, hours, search window and depth, its risk per trade and contract cap, its rules for searching again, choosing and stopping, its daily limits, its state and cycle number. And a fixed word or category (such as “TradingView signed out”, never the text of an error) when a bot is added, live, turned off, removed or in trouble, when a job is created, deleted, running, holding, paused, waiting or needs you, when a search fails, and when a daily limit stops trading or is lifted, with the amounts. Also the ID or name, the app version, your system (Mac, Windows or Linux) and when it last checked in.
• Never sent: individual trades (their times, prices or sizes), your computer’s name, account names, ids, balances or logins, what you called a bot or a job, error messages or any other text AutoBot writes. The database, like any website, can see your internet address.
• Reset AutoBot starts everything from its defaults, which includes this being on. Its question offers to delete what was sent first.
• The About row now says what leaves the computer: the above, the community sharing if you switched it on, and what TradingView and TradersPost need.

The trading-hours blocks are in the open in Autonomous

In Autonomous > New job > What to search, the hours the job trades in (Any time, Asia, Europe, New York, Custom hours, all New York time) are now in the open under the timeframe as When should it trade?, the same blocks as Find a Bot, with the switch that closes trades before the prop-firm cutoff beside them. They were already in the form, but folded away under “More: trading hours and candles”, so they were not seen. Only the candles stay under More: candles. A job whose bot already trades a chart keeps that chart’s hours, as before.

A hand Turn Off on a job that is only waiting holds it until Resume

When you turn off, by hand, the bot of an Autonomous job that was only waiting (for example its first Go Live held back by an account that is stopped for the day), the job is paused and its Go Live waits for Resume. Before, the pause let that Go Live carry on, so the bot would have gone live again the moment the account was free, and meanwhile the waiting job kept other jobs and your own searches out of Chrome (“is changing its bot”). It is now Paused on its card, nothing waits on it, and Resume goes live from where it was. A plain Pause is unchanged: a swap that has started still finishes, so a bot is never left half changed. Found on a real job before it did anything.

What did not change

Bots, alerts, limits, results and saved settings carry over. The community sharing is exactly as in 2.13.0: off until you switch it on.

Checked

• The automated tests: 4,263 run, 4,256 pass, none fail (7 are skipped by design, as before). Each new rule was cut out of the code in turn and a test failed every time (more than a hundred cut-outs), except two that change nothing (a setting that is synced when the sender starts). A second pair of eyes went over the whole feature before release; what it found (the Home first-run card and a few pages of public copy still said nothing is shared; “off” could be read as on from a damaged settings file; a row after every trade could be read as the trade; totals wiped to zero by a restart; the notice said less than was sent; and more) is fixed and tested.
• The categories are held to the app’s real sentences: every fault the deploy engine can report goes through the app’s own wording and must land in its fixed word, and every reason a job can wait or need you is driven through the real engine and must too.
• The real database: every kind of row the app builds was posted to the live private tables from a throwaway install and read back; nobody else can read them (no header and another install’s header show nothing, the three summary views refuse the public key), a row that breaks a check is refused, a row for someone else’s install is refused, and Delete my data removed the install and everything of it.
• The hold, on the real job it was found on: its account’s daily limit came back on at 6 pm New York, and the bot stayed off and the job still reads Paused. The earlier behaviour, run in the engine’s tests, put the bot live at that moment.
• A real copy of the app (its own folder, a stand-in database, started the way the installed app starts), 44 checks: on at the first start with nothing asking, the notice in Settings and in About and nothing on the Home card saying nothing is shared, the install row with the right headers in the private table and no public table touched, the computer’s name in nothing that was sent, a name change, switching off (nothing sent, still off after a restart), switching on again (the same ID), a damaged settings file starting OFF with no new ID, Delete my data, and Reset starting again on the defaults with a new ID. The Settings row and its two sheets were looked at on screen.

Not run

• A real bot’s totals sent from a real account to the real database. The totals were built and sent from realistic snapshots, not from a bot trading live.
• Windows and Linux. Run on macOS only.

---

Unsigned build.

Windows first install:

1. Run AutoBotDeploy-Setup-2.13.1.exe.
2. SmartScreen will warn because the build is unsigned — More info → Run anyway.

macOS first install:

1. Pick the right DMG — Apple Silicon = *-arm64.dmg, Intel = the one without -arm64.
2. Drag AutoBotDeploy into Applications (do not run from the DMG).
3. Open it once. When macOS says it cannot be opened: System Settings → Privacy & Security → scroll down → Open Anyway, then open it again.
4. If it still refuses: Terminal → xattr -cr /Applications/AutoBotDeploy.app then open it again.

Updates download and verify automatically; install them yourself when flat.

v2.13.0

Share your results with the community (optional), and a Reset AutoBot button

Both were asked for on 6 October 2026 and both live in Settings. No strategy changes: there is no new strategy version.

Share your results with the community (optional)

• Off until you switch it on. Settings > Community. Nothing is shared before then, and nothing is sent in the sample-data preview.
• You agree first. Switching it on opens a sheet that lists everything shared, says what is never shared, and asks for your public name: a random alias ("Calm Heron 7F3A"), your computer's name if you choose it, or one you type. The name you see is the name sent.
• What is shared: each closed trade (market, long or short, contracts, time in and out, prices, profit), the five settings of the bot that made it (Width, Period, Target, Stop, Trend), the timeframe, hours, risk and contract cap, and whether the account was paper or live; your Autonomous jobs by what they do, and what each cycle chose with what the replaced settings made live; the searches you run and their best result; bots you add, switch on, switch off or remove; daily-limit stops with the amounts. It goes to a public database, public 30 minutes after it happens, and it is self-reported: nobody checks it.
• What is never shared: your name, email or location in anything it sends, account names, ids, balances or logins, and what you called a bot or a job (they go by market, timeframe and hours). Account ids and trade ids at TradersPost are not sent either, and the reason text of a cycle is replaced by fixed words, because the app's own messages can name an account.
• Only what happens after you agree, unless you tick Also share what AutoBot already holds (off): the trades of the bots live now, finished cycles and saved searches, once.
• You stay in control. The row says who you share as, how much has been sent, how much waits, and when something failed (AutoBot keeps trying and loses nothing). Change name, Send now, and Delete my shared data, which removes everything you shared from the database and switches sharing off. It needs no account: your computer holds the key. If the database cannot be reached it is kept and tried until it works. Switching sharing off stops everything at once; if what is shared ever changes, nothing is sent until you have read the new list and agreed again.
• The public page that shows these results is not built yet. What you share is stored, and shown there once it is.

Reset AutoBot

• Settings > Reset > Reset... deletes everything AutoBot has saved on this computer (bots, results, searches, Autonomous jobs, settings, saved TradersPost accounts) and starts again as new. Charts AutoBot made in your TradingView account stay there, and your TradingView and TradersPost accounts are not touched.
• Turn every bot off first. While a bot is live, or may be (a Go Live that was cut off), or an Autonomous job is changing its bot's settings, or something is going live, turning off or searching, the button is off and says why: a reset would leave an alert trading with nothing watching its daily limit.
• Two ticks. Also sign me out of TradingView and TradersPost (on): deletes AutoBot's own Chrome profile, so you sign in again; your normal Chrome is never touched. Also delete what I shared with the community (shown if you ever shared): removes it from the database first, and if that cannot be done, nothing is reset. Left unticked, what you shared stays public and can no longer be deleted from AutoBot, because the key to it is deleted with the data.
• Type RESET to confirm. AutoBot closes and reopens, and empties its folder as it starts, before anything has loaded. It empties only its own folder.

A cut-off Go Live card says how to take it down

2.12.9 put a Turn Off button on a job's bot whose Go Live was cut off. Its red words still only pointed at Open Autonomous, so the card now says under them: no alert in TradingView? Turn Off takes down anything left and pauses the job; Resume on the job goes live again.

What did not change

Bots, alerts, limits, results and saved settings carry over, and nothing is sent anywhere unless you switch sharing on.

Checked

• The automated tests: 4,152 run, 4,145 pass, none fail (7 are skipped by design, as before). Each new rule was cut out of the code in turn and a test failed every time (more than a hundred times). A second pair of eyes went over the whole feature before release; what it found (a Reset that would have reopened the installed app on the wrong folder, trades from before you agreed being sent, account names in error text, a few ways something could still be sent after sharing was switched off) is fixed and tested.
• The real database: every kind of row the app builds was posted to the live community database from a throwaway install, read back, and the install deleted.
• A real copy of the app (its own folder, a stand-in database, started the way the installed app starts): switching on, the install row with the right headers, Delete my shared data, and Reset with the relaunch, which emptied the folder and nothing else.
• Seen live on the real app (restarted onto the build just before the card hint, with three bots live): Community is off, Reset is off and says "3 bots are live or may be", and a kill signal closed the app with no question.

Not run

• A real trade of a real bot sent to the real community database. Sharing has not been switched on with real trades.
• Reset on a real data folder. It is refused while bots are live, which they are.
• Windows and Linux. Reset retries a locked file for about ten seconds, which matters on Windows; it has been run on macOS only.
• The public page that shows what is shared is not live yet.

---

Unsigned build.

Windows first install:

1. Run AutoBotDeploy-Setup-2.13.0.exe.
2. SmartScreen will warn because the build is unsigned — More info → Run anyway.

macOS first install:

1. Pick the right DMG — Apple Silicon = *-arm64.dmg, Intel = the one without -arm64.
2. Drag AutoBotDeploy into Applications (do not run from the DMG).
3. Open it once. When macOS says it cannot be opened: System Settings → Privacy & Security → scroll down → Open Anyway, then open it again.
4. If it still refuses: Terminal → xattr -cr /Applications/AutoBotDeploy.app then open it again.

Updates download and verify automatically; install them yourself when flat.

v2.12.9

AutoBot asks before it closes, a bot stopped for the day can be resumed from its card, a cut-off Go Live can be taken down while its job waits, and "today" says which day it counts

All four came out of 6 October 2026: an app found closed in the middle of a Go Live, a job that had reached its daily target and could not be told to trade again, and a job whose cut-off Go Live could be neither turned off nor deleted for hours. No strategy changes: there is no new strategy version.

AutoBot asks before it closes

• What was wrong. Nothing stood between Cmd+Q (or, on Windows and Linux, the window's close button) and the end of the app. One was closed in the middle of a Go Live, at the step that makes the alert: the app could not tell whether an alert had been made, and nothing watched the bot or its account's daily limit until the app was opened again.
• What happens now. If a Go Live or a Turn Off is mid-way, AutoBot says so ("AutoBot is going live") and asks: Finish first, then close (the default: it waits, and closes by itself when the work ends), Close now, or Keep AutoBot open. If nothing is mid-way but bots are live, it says "2 bots are live. They keep trading while AutoBot is closed. Your daily limits are not watched, and no Autonomous job can switch settings, until AutoBot is open again." with Keep AutoBot open (the default) or Close anyway. With nothing live it closes at once, as before.
• Not asked: when the computer shuts down or restarts, when AutoBot is stopped from outside (a kill signal), and in the sample-data preview. A crash or a power cut can still leave a Go Live cut off; opening AutoBot again puts it right, as before (the bot shows "an alert may be live", it is watched, and an Autonomous job tries going live again by itself).

A cut-off Go Live can be taken down while its job waits

• What was wrong. A job whose first Go Live was cut off after the alert could exist (the app was closed), and whose account was then stopped for the day by its daily loss limit, waited for hours. Its bot's card said "Alert may be live" (there was no alert on TradingView) and offered only Open Autonomous and Details. The job's card had no Stop its bot (that needs a live bot), and Turn Off and Delete were both refused with "it is changing its bot right now ... in a minute or two", though nothing was being changed: the job was only waiting.
• What happens now. A job that is only waiting (its card says Waiting, it has no action of its own in flight, and its bot is not live and has nothing going on) can be taken over. Its card offers Stop its bot for a bot whose Go Live was cut off, and says what that does: it takes down any alert it left and pauses the job, and Resume then goes live again from where it was. The bot's card in My Bots shows Turn Off while an alert may be live, and says that the bot belongs to the job and that turning it off pauses the job. Delete job works too, with the bot (taken down and removed) or without it.
• Still left to finish: a job that is really changing its bot (a Go Live or a Turn Off in flight, an action of its own running, a live bot whose swap has asked for its Turn Off, or a job about to carry on by itself) keeps its "stop when that has finished, in a minute or two" answer.

A bot stopped for the day can be resumed from its card

• What was wrong. A job that reaches its daily profit target (or its loss lockout, or three losing cycles in a row) switches its bot off for the rest of the trading day. Its card showed Pause, its bot's card in My Bots said "Paused" and had only Open Autonomous, and nothing offered to trade again, though Resume has always lifted the stop.
• What happens now. The job's card shows Resume while a stop for the day is in force, and its bot's card shows the same Resume. Pressing it asks first ("Trade again today?"): it lifts the stop, counts the day again from zero (after a target or a lockout, so the limit then applies to what it makes or loses from now), and switches the bot back on. A job that is only paused, or has stopped itself, resumes at the press, as before.
• If the account's own daily limit stopped the bot (Settings, Daily limits), its card has Change limit: raising or clearing the limit switches its bots back on.

"Today" says which day it counts

• What was seen. A job's card said "Target reached: +$1,843 of $1,500 today" while its bot's card said "Today +$814". The figures agree with each other, but each counts something different. The sidebar's Today is the account's trading day: from 6 pm New York time, carried across TradersPost's own midnight restart of its Day's P&L. The job counts every trade it made that trading day, whichever settings made it: a swap does not start the day again. A bot's own Today shows only the trades since that bot went live. TradersPost's page restarts at midnight New York time, so in the evening it can show less.
• What changed. The sidebar's Today now says "since 6 pm ET" under its label, as Home and a bot's card do, and its tooltip says why it can differ from TradersPost's page. The job's target line has a tooltip that says what it counts. The explanations under the target and lockout switches on the setup screen now say which day they count. No figure changed.

What did not change

Bots, alerts, limits, results and saved settings carry over. When a job stops for the day, and what lifts the stop, are as they were: only where Resume is offered has changed. How the day is counted is unchanged. A job that is really changing its bot is still left to finish.

Checked

• The automated tests: 4,043 run, 4,036 pass, none fail (7 are skipped by design, as before). The new ones run the question's every path with a fake dialog and a fake timer, the real Autonomous engine (a target and a lockout reached, then Resume; a cut-off Go Live while the account is stopped, then Turn Off and Delete), and the real wording. Each new rule was cut out of the code in turn (80 times) and a test failed every time.
• A kill signal and the question, in a real Electron: a test starts a real Electron with the real question wired in, three bots live, and sends it SIGTERM. A kill signal is handled once the app is ready and closes it without asking; the same signal handled any earlier is replaced by Electron's own and reaches the question (found on the first restart of the live app with the question in it, and fixed before this release).
• Seen live on the real app: the question with three bots live ("3 bots are live"), Keep AutoBot open left it running, Close anyway closed it; the sidebar's "Today · since 6 pm ET".

Not run

• Finish first with a real Go Live in flight was run only with a fake dialog and timer.
• The Windows and Linux window-close question has been run on macOS only.
• Resume and Change limit on a real stopped bot were not pressed: they are tested on the real Autonomous engine with simulated trades and by reading the window's code.
• Turn Off and Stop its bot on a real cut-off Go Live are tested through the app's own handlers with a fake TradingView and TradersPost, and by reading the window's code.
• The figures above (+$1,843 against +$814) are explained by the evening session counting toward the next trading day. That reading fits the numbers and has not been checked against that job's own trade times.

---

Unsigned build.

Windows first install:

1. Run AutoBotDeploy-Setup-2.12.9.exe.
2. SmartScreen will warn because the build is unsigned — More info → Run anyway.

macOS first install:

1. Pick the right DMG — Apple Silicon = *-arm64.dmg, Intel = the one without -arm64.
2. Drag AutoBotDeploy into Applications (do not run from the DMG).
3. Open it once. When macOS says it cannot be opened: System Settings → Privacy & Security → scroll down → Open Anyway, then open it again.
4. If it still refuses: Terminal → xattr -cr /Applications/AutoBotDeploy.app then open it again.

Updates download and verify automatically; install them yourself when flat.

v2.12.8

Autonomous mode: an edited job follows its chart, Chrome opens by itself when a job needs it, a job starts only on a winning run, and it turns its bot off and waits for you when it cannot find new settings

All four came out of one running job on 6 October 2026. No strategy changes: there is no new strategy version.

An edited job follows its chart

• What was wrong. A job with no bot that was moved from MNQ on 5 minutes to MES on 1 minute kept everything about the chart it had left. The card still said "MNQ · 5 min"
  over a subtitle that said "MES · 1 min", its new search was labelled with the old name in Find a Bot and in Past searches, it waited out a retry that had been timed for MNQ (you
  had to press Search now to get past it), and its card went on showing the old chart's "Nothing qualified" as its last decision beside a search of another market.
• What happens now. A job is called by its chart ("MES · 1 min"), so the name follows it. A job with no bot that is moved to another market, timeframe, hours or candles
  starts afresh on the new chart: its first search is next (within a minute), and its card says nothing of the old chart's last decision or scorecard. Its history stays,
  each cycle under the name it ran with. A day's stop it is under, a pause, its record and its history are never touched by an edit. A job that has a bot, or is in the middle of
  a search, still cannot change chart.
• A job the old build had already left with the wrong name is put right the first time the app starts on this one, and the log says so.
• Found on the way: Save and start on a job that had never been started only saved it. It now starts it, as it does for a new job. The toast after a save says where the job is.

Chrome opens by itself when work is waiting for it, and only a closed Chrome

• What was wrong. A job that was ready to search found Chrome closed, said "Waiting · Chrome is closed", and waited for you to press Open Chrome. After Search now, nothing happened.
• What happens now. A job that is ready to search, or to change its bot, opens Chrome itself when it is closed (it was seen to open it in two seconds and start the search). Only a closed Chrome:
  one that is open (busy, or signed out of TradingView or TradersPost) or one that is starting is left alone, and so is a search's own Chrome. A try that did not work says
  "Chrome would not open" on the card and in a notification with the reason, and it tries again in five minutes, or at once when you press Search now or Resume.
• The same for the other places that were waiting for you: a Turn Off that is waiting to finish, Pause all, Resume all, a daily limit's reset that still has bots to switch back on, and an
  Autonomous job's live bot (its trades are read through Chrome); and the buttons that told you to "Press Open Chrome on Home or Settings, then try again" (Open chart, Refresh on a bot's trades,
  Validate, Check now), which now open it first and then do what you pressed.
• Left as it was, on purpose: a daily limit that cannot be checked because Chrome is closed is still not reopened by itself (that is your earlier decision); the app says so as loudly as it can.
  Housekeeping nobody is waiting for (the chart sweep, the hourly health check on its own schedule, the support file) never opens Chrome.

A job starts only on a winning run

• What was wrong. A job picked settings whose curve had topped out and turned over: score 70, +$6,243, profit factor 2.50 over 42 trades, and a last trade of +$45 after a full -$500
  loss. The search already read each test's last trade and that "win" passed. Nothing looked at where the test ended.
• What happens now. A result counts only if its last 3 trades in its test all made money (a switch under Which result counts: Start on a winning run, 1 to 5 trades, on at 3).
  A trade that made exactly $0 is not a win, and trades the app could not read count as not sure, so a result it could not read is left out too. Only where the test ends counts: older
  losses do not. A job saved before this has it on. The card says why a result was left out ("its last 3 trades were not all wins (newest first: +$45, -$500, +$400)") and the pick
  says the run it starts on ("its last 3 trades all won"). The search reads the last five trades of every test (it read only the last one).
• Not a forecast. It filters recent form; it narrows the choice (with a 60% hit rate roughly one result in five is on a run of three), so on a quiet day nothing may count, and the job then
  holds or keeps what it has. It is read when the search ends: a swap that has to wait for an open trade to close can start a little later than the test it was chosen on.

A job turns its bot off and waits for you when it cannot find new settings

• What was wrong. After the losses a job is set to (2 in a row), it searches. If nothing qualified, the default was to keep the bot trading the settings that had just lost; the other choice
  turned the bot off but looked again by itself every half hour and could put a bot live with nobody there.
• What happens now. When nothing counts has a third choice, Turn the bot off and wait for me, and it is the default. The bot is switched off (never with a trade open: it waits
  for the account to be flat), the job stops and needs you (a notification that shows even with notifications off), and it stays off until you press Resume, which searches again and puts
  the best result live. A job you already set to keep its settings, or to turn the bot off and keep looking, keeps that choice; a new job, and one that said nothing, has the new one.
  With no bot there is nothing to switch off (the job holds and tries again, as before), and an alert about to expire is still renewed.

What did not change

Bots, alerts, limits, results and saved settings carry over. A job's search, its swap and its Go Live are the steps they were. How a bot you made yourself trades is unchanged: only the buttons named above open Chrome for you.

Checked

• The automated tests: 3,995 pass, none fail (7 are skipped by design, as before). The new ones run the real Autonomous engine (a hold, a move to another chart, a stop, a Resume), the real file
  (a stale name read back and put right, the stop kept across a restart), the real words, the real functions that open Chrome (against a fake clock: only a closed Chrome, the cooldown, a press, a
  launch that fails), the parse of the List of trades on a real fixture, and the setup screen's form. Each guard was cut out of the code in turn and a test failed every time.
• Seen live on the real app: a job whose Chrome was closed opened it and started its search; the repaired job reads "MES · 1 min" with its own last decision; the setup screen shows
  Start on a winning run.

Not run

• A job stopping after two real losses with nothing to switch to, and its Resume, have been run only in the simulation: they need two real losses and a search that finds nothing.
• A search under the winning-run rule has not finished yet on the real app (the first one will say what it left out and why).
• A job moved to another chart on the real app was checked on a copy of the real file and in the tests, not by moving your job (that would have started an hour-long search).
• Turn Off, Pause all and the four buttons opening a closed Chrome are tested against fakes, not pressed with Chrome closed.
• Known limits. The winning-run check is made when the search ends and is not made again at Go Live. A +$45 scratch counts as a win. A job's chart still cannot change once it has a bot.

---

Unsigned build.

Windows first install:

1. Run AutoBotDeploy-Setup-2.12.8.exe.
2. SmartScreen will warn because the build is unsigned — More info → Run anyway.

macOS first install:

1. Pick the right DMG — Apple Silicon = *-arm64.dmg, Intel = the one without -arm64.
2. Drag AutoBotDeploy into Applications (do not run from the DMG).
3. Open it once. When macOS says it cannot be opened: System Settings → Privacy & Security → scroll down → Open Anyway, then open it again.
4. If it still refuses: Terminal → xattr -cr /Applications/AutoBotDeploy.app then open it again.

Updates download and verify automatically; install them yourself when flat.