Building KryllOS, the future of trading bots. Open-source, self-hosted & no-code trading automation solution.

Joined November 2017
Pinned Tweet
We are very proud to announce Kryll Open System (KryllOS), your personal automated trading system. Hosted on your machine or your VPS. Your keys. Your bots. Your rules. Build your strategies with the drag-and-drop editor you know, or let AI build them for you. Describe what you want in plain language, and watch your strategy take shape. KryllOS isn't just another automated trading tool, it's a complete system, an open playground. Create your own editor blocks. Your own apps. Your own AI agent skills. Share them on the marketplace and install those from the community. One more thing : KryllOS will be free and open source. If you can dream it, you can build it.
15
24
1
75
8,801
Every trading strategy should walk the same path: backtest, paper trading, then real money. Step 2 is where you find out a strategy only looked good on old data. KryllOS is built around that path: no code, open source, on your VPS.
2
1
5
504
We gave 4 trading bots the worst ideas we could find. Starting today, they’re paper trading Bitcoin on KryllOS. Meet the cast and pick your champion in the replies.
2
4
663
🎰 The Gambler flips a coin every hour. Heads, it bets. After every loss, it doubles the next bet. That’s the classic casino martingale.
1
1
52
Same rules for all four: BTC/USDT spot, 10,000 virtual USDT each. Each bot is built from blocks in the KryllOS trading strategy editor. See you in a few days for the first results!
50
It works on paper
🤖 Made with AI
Until today, your trading strategy only lived in the past. Paper trading is now live on KryllOS. Real market, virtual money.
5
541
Backtest: how your trading strategy would have done. Paper trading: how it does now, on the live market, with virtual money. The first kills bad ideas fast. The second catches the ones that only looked good on old data.
1
703
A trading strategy can lose almost half its capital without a single losing trade. The only cost: a 0.1% exchange fee on every buy and every sell. A backtest without fees is easy mode, then a wall on the real market. The KryllOS backtest engine simulates them on every order.
1
1
2
629
Same trading strategy twice. One with a stop loss, one without. No stop: +157%, max drawdown 48%, worst trade -19.5%. Stop at -7.5%: +236%, max drawdown 40%, worst trade -8.3%.
1
2
739
The stop fired only 10 times in 121 trades. It cut 5 losses by up to 12 points, and cost 2.2 points at most on the other 5. Every other trade was the same. Built and backtested in the KryllOS trading strategy editor.
2
87
Overfitting is one way a backtest lies. A lazy engine is the other. A lazy engine only looks at where each candle closes. KryllOS turns every 1-minute candle into up to 16 ticks that run through its open, high, low and close, and checks your stops and targets at each one. On Oct 10, our stop was hit inside a candle, on tick 11. A close-only engine would have filled it a minute later, $362 lower. In our backtest, 24 of 61 orders were filled inside a candle like this. Find out where your stops land before real money does. KryllOS gives you this engine for free: build your trading strategy in blocks, no code, and backtest it tick by tick.
Backtests lie. Quantopian measured by how much. They froze the code of 888 trading algorithms, then ran them on 6+ months of market data they had never seen. The backtest Sharpe ratio, the number everyone screenshots, explained less than 2.5% of how they did next.
3
3
726
Backtests lie. Quantopian measured by how much. They froze the code of 888 trading algorithms, then ran them on 6+ months of market data they had never seen. The backtest Sharpe ratio, the number everyone screenshots, explained less than 2.5% of how they did next.
1
1
8
1,004
The fix is boring: run the strategy on periods it has never seen, and lots of them. That's the part KryllOS makes easy. You build your trading bot without writing code, and Multi-Backtest runs dozens of backtests in parallel on your VPS. Free and open source.
1
113