Questions.

You pick a trading strategy — one built by us, or one you build yourself — and it runs around the clock on your own Kalshi account. You watch what it does from a dashboard we call the cockpit, and you set the limits it has to trade inside.

What we sell is software. We are not a fund, we do not pool anyone’s money, and we do not manage accounts. The way we put it internally: we deploy software to other people; we don’t manage their money.

The other half is the marketplace. Someone who can describe a good trading idea can build it here, prove it out on recorded and paper markets, and eventually sell it to people who would rather buy than build.

A quant strategy is a set of conditions written down precisely enough that a computer can check them. When the conditions are met it acts; when they are not, it does nothing. That is the whole idea. The discipline is in the "does nothing" half, which is the half a human is worst at.

The number that surprises people is the ratio. A bot looks at every market it is allowed to look at, every hour, and reads the order book many times a minute. Out of tens of thousands of observations in a day, a strategy might place a handful of orders. Almost everything it sees, it passes on.

That is not the bot being lazy. Deciding whether to participate at all is the part of trading that matters most — a strategy that is choosy about when it shows up is doing its main job, and an hour it sits out is an hour with nothing of yours at risk.

A Kalshi contract asks a question with a definite answer — will bitcoin be above a given price at the top of the hour, say. You buy the yes side or the no side. Prices run from 1 to 99 cents and read like a probability. At expiry the contract settles at $1 if the answer was yes and $0 if it was no.

Kalshi is a designated contract market regulated by the Commodity Futures Trading Commission. You open your account with Kalshi directly, in your own name, and you fund it yourself. We are not part of that relationship and we never see your Kalshi login.

Parachute trades the crypto markets on Kalshi. Your account and your money stay with Kalshi throughout.

Most people who sign up buy a strategy someone else built, set their limits, and watch. They never open an editor. That path is designed to work on its own.

The other path is for people with an idea. You describe it, an AI turns it into a strategy on our harness, you test it against recorded markets, run it on paper, and only then decide whether to put real money behind it.

You do not have to choose up front. Buying first and building later is the normal order.

Running a bought strategy is a purchase, a key, and a few limits. There is nothing to write.

Building is where people expect to need a maths background, and this is the part that has genuinely changed. You describe the idea in plain language to an AI you already use, and it does the work against our harness: writing the strategy, running it over recorded markets, and telling you what it found. You read the results and decide.

What you do need is judgement — about whether an idea makes sense, and whether the evidence for it is real. That is not a maths skill.

Nothing is shared with another customer. Your strategy runs on its own instance, against your account, with your limits.

The strategy

The algorithm itself, running around the clock inside the hours and limits you set.

Your own instance

A dedicated machine running your bot. You do not install or host anything, and you do not share it.

The cockpit

Your limits, the kill switch, the trading schedule, and the live state of the bot — what it is looking at and what it decided.

The visual record

Every hour the bot traded, as one square: green earned, orange lost, blank sat out. Plus the trades themselves, in a list.

A strategy that has to wait for you cannot act on a condition that lasts ninety seconds at three in the morning. So the bot acts on its own, and your control is exercised in advance and in aggregate rather than trade by trade.

That is what the money envelope is for: the most you want at risk in one market, the most you want at risk overall, the most you are willing to lose in a day, and the hours the strategy is allowed to trade at all. Inside that envelope the bot acts; at its edges it stops.

And you always keep the two immediate controls: the kill switch here, and your own Kalshi account, where you can cancel or close anything yourself.

Your trading balance lives in your own Kalshi account. You opened it, it is in your name, you funded it from your own bank, and you can withdraw from it whenever you like without asking us. We never take custody of a dollar.

This is the single most important thing about how Parachute is built, and everything else follows from it. We are not a fund and we do not pool money. There is no Parachute balance, no deposit, and no lock-up.

What we have is permission to trade on your account, which you granted by creating an API key and pasting it in, and which you can take away at any time.

Cash you have not put to work sits as your Kalshi balance. Money the strategy has committed is held by Kalshi as collateral against the contracts you hold, exactly as it would be if you had bought them by hand.

You can log into Kalshi at any time — without us, in a different browser, on your phone — and see every position and every order. Nothing about running a bot changes your access to your own account.

What we hold is an API key that you created inside your own Kalshi account. We use it to read your positions and balance and to place and cancel orders on the markets your strategy trades — that is the whole of what our software asks it to do. We do not have your Kalshi password, your email login, your two-factor device, or any banking detail of yours, and we never move money to ourselves.

There is one kind of movement the software does perform, and we would rather name it than have you find it: at your instruction it can shift collateral between the separate balances Kalshi keeps inside your one account, so a strategy can trade where it needs to. That is money moving within your account — never out of it, and never to us.

Withdrawals are something you do yourself, signed in to Kalshi. And if you would rather we could not act at all, delete the key in your Kalshi account: it stops working at Kalshi’s end, without needing our cooperation.

If we vanished overnight, your balance and your open positions would be exactly where they were. Open contracts would settle at expiry on Kalshi’s schedule as normal. You would revoke the API key, and that would be the end of your relationship with us.

This is the practical difference between non-custodial software and a fund. There is nothing to redeem, no queue to join, and no administrator to wait for — because we were never holding anything.

Your Kalshi private key is encrypted at rest. Only server-side code decrypts it, and only to sign a request to Kalshi. It is never included in a page and never in an API response to your browser — once saved, it is write-only: nothing can read it back out through the dashboard, including you.

The signing itself happens server-side too: every request to Kalshi is signed with your key at the moment it is sent, rather than the key being handed to anything else.

You are not dependent on us getting this right forever, which is the more important point. You can rotate or delete the key in your own Kalshi account whenever you want, and a deleted key is dead immediately whatever anyone else holds.

Orders the bot places are ordinary orders on your account. A resting order can be cancelled and a position can be closed by you, directly in Kalshi, without going through us and without our permission.

This is worth knowing before you need it. It means you are never waiting on our dashboard, our support, or our servers being up to get out of something.

We read your positions live from Kalshi whenever you look at them, and we store your fills, settlements and balance so the dashboard can show you your record without hammering the exchange. That is what makes the hourly grid and the trade history possible.

We do not see your bank, your Kalshi login, your identity documents, or anything you do on Kalshi outside the markets we read.

The envelope is set once and binds everything the bot does, whatever strategy is running. Three of the four are always in force — you can raise or lower them, but you cannot switch them off. Only the profit lock can be set to zero to disable it, because "stop when you are up" is a preference in a way that "cap my exposure" is not.

Changing any of them is a direct edit that the bot picks up on its next cycle. There is nothing to redeploy.

Max position size

The most the bot may hold in any one market. It binds cumulatively and at the worst-case price it could be filled at, so the bot cannot creep past your number by adding to a position it already has.

Max exposure

The most the bot may have at risk across every market at once.

Daily loss limit

The most you are willing to be down in a day. Reach it and the bot stops trading for the rest of the day.

Daily profit lock

The other end of the same idea: once the day is up by your number, stop and keep it. Set it to zero to turn it off.

Trading schedule

Which hours the strategy may trade at all. Painted hour by hour, per strategy — an hour you leave off is an hour nothing happens.

Kill switch

Stops the bot entering anything. It is at the top of every screen in the dashboard, not buried in a settings page, because you should never have to go looking for it.

The bot picks the change up on its next loop, so it finishes the loop it is in and places nothing after that. It sits at the top of every screen and it is a single explicit state, not a strategy setting — the "I do not want to trade today" control. Nothing the algorithm does can override it.

It is deliberately blunt and deliberately outside the trading loop, and it is the one control we would rather be simple than clever.

Positions you already hold are not conjured away by it — and any take-profit or stop-loss the strategy was holding stops too, so a killed bot is not watching them for you either. That distinction matters, and we would rather you learned it here than in the moment — the next answer is about exactly that.

Contracts you hold when you hit the switch remain in your Kalshi account. Left alone, they settle at their expiry on Kalshi’s schedule, at $1 or $0 per contract like any other contract.

If you want out of them sooner, close them yourself in your Kalshi account — it is your position and you do not need us to do it.

We think a stop that silently dumped your positions into whatever price was on screen would be the more dangerous behaviour, and it is not what "stop" should mean. So stop means stop starting anything new, and getting out is a decision you make deliberately.

A safety control that fails open is worse than no control, because you were relying on it. So the direction of failure is fixed: if the bot cannot establish what your caps are, or cannot establish what you are currently holding, it does not enter anything.

The same principle runs through the rest of it. A strategy the platform cannot fully resolve is allowed to watch the market and not to trade it, rather than being allowed to trade on a guess.

Caps, the daily loss limit, the profit lock, the schedule and the kill switch are all live settings. You do not need to redeploy anything or wait for a window.

Money controls are also always yours, whatever else is true of your account — they are tied to being the account owner and nothing else. As the owner, there is no state in which you can see the dashboard but cannot reach your own kill switch. (A viewer you invited can see it and cannot flip it; that is the point of a viewer.)

Crypto markets do not close, which means "always on" is a choice rather than a default worth accepting. Some people run a strategy only in the hours they understand; some run it while they sleep and not while they are watching.

The grid is a full week, seven days by twenty-four hours, and you can read it in Eastern time or UTC. Each hour names one strategy or it is off — so if you own several, different hours can belong to different ones.

Below the hour there is a finer control: each strategy has an entry window, a pair of handles that say which minutes inside an hour it may open a position in.

There is no application, no sales call, no approval queue and no waiting list. Checkout comes first, and everything after it is the wizard below.

The one step whose speed is not ours is Kalshi’s: you need a funded Kalshi account before you start, and opening one is their process, not ours. If you already have one with a balance, the rest is a single sitting.

Buy

One monthly subscription, by card. This happens before the wizard, and it is what creates your account.

Terms

Read and accept them. They are short, and they say we never hold your money.

Password

Set one for your Parachute account.

API key

Create a key inside your own Kalshi account and paste it in. We never ask for your Kalshi password.

Verify

We check your Kalshi balance clears the $200 minimum, and use what we measure to seed your four limits with sensible starting numbers.

Strategy

Choose which strategy runs on your account.

Limits and schedule

Adjust the seeded limits to what you actually want, and paint the hours it may trade.

Launch

The bot goes live and you land in the cockpit.

The Kalshi account has to exist first, because the API key comes from inside it. Opening and funding it is done directly with Kalshi.

Setup asks for that key, and nothing else about your Kalshi account. We never ask for your Kalshi password.

The minimum is not a deposit and it is not paid to us. It sits in your Kalshi account, and it exists because a balance below it cannot carry a position large enough for a strategy to behave the way it was designed to.

It is deliberately low — this is meant to be something you can try without it being a big commitment. If you want to start under it, write to us: there is a route that pays a couple of months up front instead, and we would rather tell you it exists than have you bounce off the number.

The subscription starts at $100 a month and covers the strategy, the machine it runs on, the cockpit and the record.

The $200 is a minimum balance, not a fee: it is your trading money, it stays in your Kalshi account, and you can withdraw it.

A card is required to get past checkout, and you can cancel from your settings whenever you want.

We do not take a percentage of what you make, and we could not: we never hold your money, so there is nothing for us to take it out of. The subscription is the whole of what you pay us.

Kalshi charges its own fees on trades, as an exchange does. Those come off at Kalshi’s end, and every profit-and-loss figure shown in the cockpit is already net of them — we do not quote a number that quietly excludes the fees you paid.

Cancelling stops the next charge and you keep the subscription for the period you have already paid for. There is no notice period, no minimum term and no exit call.

You do not have to cancel to stop trading, either. The kill switch costs nothing and stops it within a loop, and turning the bot off is not the same decision as ending the subscription.

When a subscription lapses the bot is halted — it stops entering anything. Your Kalshi account, your balance and any positions you hold are entirely unaffected: they were never ours, and cancelling a software subscription does not reach them.

Your bot’s server is retired at least a day after the subscription ends. Your account and your settings are kept, not deleted, so if you come back the record you built is still there and a new server is provisioned for you.

If you would rather it were all gone, deleting your Parachute account is a button in your settings and it wipes your side. One thing to know before you press it, because it sits oddly next to the paragraph above: deleting ends the subscription immediately and does not refund the rest of the month you already paid for. If you are within a paid period and simply want to stop, cancel instead and let it run out. It does not touch your Kalshi account or the money in it either way — we could not reach those to delete them even if you asked.

The bot runs on a machine we provision for you. It keeps trading when your laptop is shut, your phone is off, and your internet is down.

Everything you touch is a web page.

These markets run continuously, they are liquid enough to get in and out of, and they settle quickly. That last part matters more than it sounds: a market that resolves within the hour gives a strategy many small independent decisions a day instead of a handful of slow ones, which is what makes a rule testable at all. It also means you are never holding something for weeks waiting to find out whether you were right.

Kalshi is the only venue the platform trades today. Other venues and other kinds of market are on the roadmap; nothing on this page assumes them, and we would rather under-promise the map than sell you a plan.

In a day the bot reads the market tens of thousands of times across every contract it is allowed to consider. Out of that, the number of orders that actually go out is typically counted in tens.

The cockpit draws that narrowing as a funnel, live, so you can see where the day went rather than infer it. Each stage throws away most of what reaches it.

A quiet day is not a broken bot. It is the most common kind of day.

Scanned

Every market the bot looked at.

Passed gates

The ones that cleared every risk and quality gate — your caps included.

Signals

The ones it actually wanted to trade. This is where the number collapses.

Attempted

Orders the bot tried to place.

Placed

Orders that reached the exchange.

Filled

Orders that got a counterparty. Not every order does.

The most common reasons, in rough order: the market is not offering what the strategy needs, the hour is outside your schedule, you have hit a cap or the daily loss limit, or the kill switch is on.

The cockpit tells you which of those it is rather than leaving you to guess, and the hourly grid shows sat-out hours as blank squares so a quiet stretch is visible instead of ambiguous. On a day with no signals at all it says so in as many words: the gates kept you out of the market, and that is protection rather than a problem.

It is worth internalising early: a strategy that trades constantly is not a better strategy. Not participating is a position.

Kalshi contracts are fully collateralised. You put up the cost of the contract, and it settles at $1 or $0. You cannot end up owing money on a position, and you cannot lose more than you put into it.

That is a property of the market rather than a promise from us, which is why we are comfortable stating it plainly. Your risk is bounded by what you have committed, and your caps are what decide how much that ever is.

Each hour on the schedule names exactly one strategy, so two never run against your account at the same moment. What you can do is hand different hours to different strategies you own, and change which one holds an hour whenever you like.

Each strategy you buy carries its own price, set by its author, so a second one is a second line on your bill rather than something bundled in.

An outage on our side means orders stop being placed. It does not touch your Kalshi balance and it does not touch positions you hold — those live at Kalshi, and open contracts settle at expiry on Kalshi’s schedule whether or not anything of ours is running.

And you keep the direct route throughout: you can sign into Kalshi and cancel or close anything yourself without us.

Two strategies can make the same amount and be nothing alike: one grinds steadily, the other lurches. The Sharpe ratio is the standard way of separating them. Higher means the same return arrived with less violence along the way.

It is quoted annualised, and it is meaningless without a duration: a spectacular Sharpe over two weeks is noise, and the number only starts to mean something once there is enough history behind it. We show it with that qualifier rather than as a bare figure.

We benchmark every strategy the same way so that they can be compared honestly. For scale: the best-known quant funds run at roughly 1.6 to 2.5 annualised, sustained over years, and that is considered exceptional.

Our range as of September 2026 runs from about −8 to about 26. Both ends are real and both should be read sceptically — a very high figure over a short window is far more likely to be a small sample than a discovery, and a negative one is a strategy that has been losing money. We would rather publish the range than the flattering end of it, because the flattering end is the thing everyone else shows you.

This is a benchmark, not a promise. It describes what we have measured across our own strategies, not what yours will do, and nothing on this page is a projection of your results.

Markets change, and a rule that fit last quarter can stop fitting. Every strategy here can lose money, including ones with a good history, and no figure anywhere on this site is a forecast.

What a record is good for is elimination. It will not tell you a strategy will work; it will often tell you one will not.

A backtest replays real recorded market data and asks what the strategy would have done. Ours runs on our own recorded scan history of the real market rather than on invented prices.

It is still a rehearsal, and here is where it stops short. It cannot know whether your order would truly have been filled at that price, or how the market would have moved because you were in it. It models when a strategy would have entered and exited — it does not model how big the position would have been, so two sizing methods backtest identically. Treat a strong backtest as a reason to keep looking, never as a result.

The honest use of a backtest is to kill ideas cheaply. Most ideas deserve it.

We would rather say this plainly than imply a paper-trading mode you will go looking for and not find. What you can do today is start small: your position cap is yours from the first day, so you can run the thing for real with an amount you would not mind losing, and watch what it actually does on your own account.

Backtests are the other half of it. They are free, they run against recorded markets, and they are genuinely useful for ruling ideas out — with the limits described a few questions up.

The engine does have a shadow mode, where it makes real decisions on live markets with nothing sent to the exchange. It runs today as an internal evidence-gathering tool rather than a screen you can open, and where paper results are shown they are labelled as paper and never mixed into your real trading record. If a proper try-before-you-fund step is what would decide it for you, tell us — that is worth knowing.

A strategy’s figures are computed from what it actually did — its deploys, its trades, its settlements — not from a claim its author typed in. Every strategy on the marketplace is benchmarked identically, so two of them can be compared without adjusting for who is telling the story.

Where the evidence is thin, the qualifier says so. A short record is shown as a short record.

We expose the platform to AI assistants over MCP, an open standard for connecting a model to tools. Your assistant can read your account context and our documentation, write and edit a strategy, and run backtests over recorded markets, then tell you what happened.

The loop is: you have a thesis, the AI grounds it and tests it, you read the evidence, and you decide. What has changed is not that computers can trade — it is that describing a strategy well enough to test it no longer requires you to write the code.

The intent is that quant trading becomes about as hard as programming has become: still a skill, no longer a priesthood.

One practical note so you are not surprised: the building tools are a separate plan from the standard subscription, which buys and runs a strategy. Selling what you build is a further step again. If building is what you are here for, say so when you write to us.

MCP is an open standard, so this is not a partnership list and we are not picking a winner for you. You connect the assistant you already pay for and already trust.

You mint a token in your settings, point your assistant at our server with it, and it can work on your account. Revoking that token cuts the connection.

This is a structural boundary rather than a promise we are asking you to trust. There is no tool over that connection for deploying a strategy, controlling the bot, touching your credentials or placing an order — they do not exist to be called, so no amount of persuading a model reaches them.

A strategy an AI edits is inert until a human opens the dashboard and clicks Deploy. That click is yours, and it is the only way a strategy change reaches a live account.

We do it this way because a conversation is a bad place to keep a money control. Chat is for figuring out what you want; the deploy button is for meaning it.

The machinery works today, end to end: build a strategy, backtest it, run it on paper, list it with its record, connect a payout account, and receive a monthly statement for what it earned. A buyer subscribes to it monthly, and if you later take the listing down, the people already running it keep it until the period they have paid for runs out.

What is deliberately not open yet is publishing to anyone who asks. We are running the first author through the whole loop ourselves before we invite others, because the first thing a creator marketplace has to get right is paying its creators correctly, and that is not a thing to find out about in public.

If you want to be in the first group of outside authors, write to us — that list is being kept.

A strategy can be sold sealed, where the buyer runs it and sees what it does but not the parameters behind it, or open, where they can see and tune it. That is the author’s choice when they record the version they sell.

Sealed means sealed in the software, not merely in the interface: a sealed strategy’s parameters are withheld from the buyer’s account by the platform itself, rather than hidden by a screen that could be worked around.

Buying, setting limits and watching is a complete way to use Parachute. The building tools are a separate part of the product and they stay out of your way if you do not ask for them.

Kalshi operates as a designated contract market under the Commodity Futures Trading Commission. You trade on it in your own name, under your own agreement with them.

Parachute sells software. We are not a broker, we do not route your orders through us as principal, and we do not take custody of funds.

We do not pool money, we do not manage accounts on anyone’s behalf, and we do not tell you what to trade. You choose a strategy, you set the limits, and it runs on your account.

Nothing on this site or in the product is a recommendation to buy or sell anything, or advice about whether any of it suits your circumstances. If you want that, the person to ask is a licensed adviser who knows your situation.

The structural differences are real: Kalshi is a CFTC-regulated market with an order book, where you trade with other participants rather than against a house, and where you can sell a position back before it resolves instead of waiting to find out.

None of which makes it safe. Prices move against people constantly, and a strategy can be wrong for a long time. If you are looking for a reason to put in money you cannot afford to lose, this is not one.

We are not going to soften this. Every strategy here can lose money, including ones with a good record. There is no arrangement in which the software is responsible for your results — you set the limits, and the outcomes are yours.

What the product gives you is control over the size of the question: caps on any one market and on your total exposure, a daily loss limit, hours you allow it to trade, and a switch that stops it. Use them.

Our risk disclosure, linked at the foot of this page, spells this out in full, and it is worth the two minutes before you sign up.

Because everything happens in your own account, Kalshi is the source of record for what you traded and what it settled at. We are not a tax adviser and nothing here is tax advice; how any of it is treated depends on where you are and who you are.

You invite them by email from your settings. A viewer can read your account — the overview, your positions, the bot’s trading operations, their own settings, and the marketplace to browse. They cannot touch your limits, your schedule, your credentials or your money controls, they cannot buy anything, and they cannot place an order.

Those controls stay with the account owner by design, and no invitation widens them. It is a read-only seat, not a junior version of yours.

The strategies we publish run on our own accounts before they run on anyone else’s, which is the only version of this we would be comfortable selling.

If you have a question this page does not answer, write to info@parachute.fund. A real person reads it.