How to choose a trading bot: Freqtrade, OctoBot, Jesse, and other open-source bots
A bot executes the rule. It does not make the money.
Short answer: an open-source bot automates orders from rules you already have. It does not decide whether the strategy makes money. Write the strategy and take the backtest seriously: Freqtrade, or Jesse if you care more about research. Configure a grid or a recurring buy with little code: OctoBot. Let a model trade from a conversation: CloddsBot or OKX's official toolkit, which is the highest-risk group. Whichever you pick, start in a simulated mode.
How to choose: if you write Python and want a real strategy, start with Freqtrade; if you care more about research and validation, choose Jesse. If you do not want to write code and only want a grid or a recurring buy, choose OctoBot. If you already use OKX and want an assistant to read prices and positions, use OKX Agent Trade Kit and start read-only. Conversational tools can misread you. If you do not have a strategy yet, do not install a bot.
Who it is for: someone who already has buy and sell rules and wants a program to send those orders without watching the screen. Python users can start with Freqtrade or Jesse. People who will not write code, and only want a grid or dollar-cost averaging, should look at OctoBot. People already on OKX who want an assistant to check prices and positions should look at OKX's toolkit, in read-only mode first.
Who it is not for: someone who does not have a strategy yet and hopes the bot will make the money. This page compares open-source programs that can send an order. It is not the backtesting-framework comparison, not the China broker-terminal comparison, and not the how-to-choose tool order. A bot executes. It does not earn.
Categories: the bots on this page fall into three groups. Strategy frameworks you code yourself: Freqtrade and Jesse. Ready-to-configure bots: OctoBot. Conversational tools that turn a sentence into an order: CloddsBot and OKX Agent Trade Kit.
The same comparison in Chinese is how to choose a trading bot (Chinese). The order of data, then a backtest, then paper trading, then an agent is a different article: how to choose AI quant tools. The nine-framework backtest comparison is how to choose a backtesting framework. China broker terminals are how to choose a quant trading platform. QMT versus PTrade, where the code runs, is how to choose between QMT and PTrade. None of those is a language version of this page.
This site has no commercial relationship with the projects below. Stars and maintenance status come from GitHub and refresh daily. The other facts were checked against each project README and its licence on 13 Oct 2026. Licences and supported exchanges can change.
01 · Five bots, one table
| Tool | What it is | Markets | Licence | GitHub stars | Maintenance |
|---|---|---|---|---|---|
| freqtrade | A crypto bot where you write the strategy, with backtesting, dry-run, and machine-learning parameter tuning | Crypto spot and futures | GPL-3.0 | 56k | Commits today |
| OctoBot | Closer to a finished product: grid, DCA, TradingView signals, and language-model connectors can be configured directly | Crypto, 15+ exchanges | GPL-3.0 | 6.9k | Commits today |
| jesse | A crypto strategy research and trading framework, focused on backtesting, parameter optimization, and live trading | Crypto spot and futures | MIT | 8.7k | Commits today |
| CloddsBot | An AI trading terminal you operate by conversation, running on your own computer | Prediction markets, crypto spot, and perpetual futures | MIT | 3k | Last commit 25 days ago |
| agent-trade-kit | OKX's official MCP server and command-line tool, so an AI assistant can use your OKX account | OKX only | MIT | 492 | Last commit 8 days ago |
Licences and markets were checked on 13 Oct 2026. The maintenance column is the live reading, not a promise that a quiet repository will stay quiet.
02 · Three categories
- Strategy frameworks you code yourself: Freqtrade and Jesse. You write the buy and sell rules in Python. The project handles the backtest, the simulation, and the exchange connection. Slower to start, and you can see what each step does.
- Ready-to-configure bots: OctoBot. Grid trading, dollar-cost averaging, and similar setups are already built. You fill in parameters and it runs. Fast to start. Parameters are not a strategy, and a careless fill still loses money.
- Conversational AI: CloddsBot and OKX Agent Trade Kit. You give an instruction in ordinary language, and a language model calls a trading API. Least work to type, and the easiest place for "it understood you wrong."
03 · Freqtrade: the strategy bot most people use
Freqtrade is a free, open-source crypto bot written in Python. It supports spot and futures, and you can control it from Telegram or its own web interface. It includes backtesting, plotting, and money-management tools, can tune strategy parameters with machine learning, and includes an adaptive prediction module called FreqAI.
Dry-run is the simulated mode: the bot runs without spending real money. The project's own README is direct. The software is for learning. Do not risk money you cannot afford to lose, and always start in dry-run. The licence is GPL-3.0. If you change the code and distribute those changes, they stay under the same licence.
04 · OctoBot: configure it, write less code
OctoBot is closer to a finished product. Grid, dollar-cost averaging (DCA), and crypto-basket strategies can be configured directly. It can turn a TradingView indicator or strategy signal into an order, and it can connect a language model on OpenAI or Ollama. It supports Binance, Coinbase, MEXC, Hyperliquid, and more than 15 exchanges. Underneath, it uses the exchange library CCXT.
It can also paper-trade with simulated funds on live data, and it includes a backtest engine. You can control it from a desktop app, a phone, the web, or Telegram. The licence is GPL-3.0.
05 · Jesse: a framework aimed at research
Jesse describes itself as a crypto trading framework, with the weight on research: multi-timeframe, multi-symbol backtests that avoid lookahead, parameter optimization with Optuna, Monte Carlo analysis, and a significance check on the rules. It ships with more than 300 technical indicators, plus paper trading and live trading. The licence is MIT, which is less restrictive than the two GPL-3.0 projects above.
It fits when you care whether the strategy actually works, more than whether it is running by this afternoon.
06 · Conversational AI: CloddsBot and OKX's toolkit
CloddsBot is an AI trading terminal that runs on your own computer. It covers prediction markets such as Polymarket and Kalshi, plus crypto spot and leveraged perpetual futures. Its README mentions leverage up to 200x. Higher leverage means a wrong call can wipe out the stake faster. Beginners should not use high leverage. The catalog entry for CloddsBot is the Chinese tool page.
OKX Agent Trade Kit is OKX's official toolkit: an MCP server and a command-line tool, so assistants such as Claude or Cursor can use your OKX account. It runs on your machine, and the API keys stay on that computer. A --read-only switch lets the assistant see prices and positions without placing orders. The companion agent-skills repository is a set of instruction files for those assistants. Both licences are MIT. The three ways to connect, where the key lives, and Cursor's tool-count limit are in how to choose OKX's AI trading tools.
The shared risk in this category: your words pass through a language model before they become an order, and that step can be wrong. Start in read-only mode, and have it repeat the action it is about to take before any order goes out.
07 · Before real money
- Simulate first. The strategy bots above all have a simulated mode. Run it for at least two weeks and see how far it sits from the backtest. The walkthrough is paper trading after a backtest. The same article in Chinese is paper trading (Chinese).
- API keys: trade permission only. Do not enable withdrawal. If the key leaks, the other party still cannot move the assets out.
- Check maintenance. Exchange APIs change. A bot that nobody maintains will eventually fail to connect. Of the bots in this catalog, Kelp and stoictradingAI have archived repositories. They are for reading code, not for trading.
- Be wary of "free arbitrage bots" and "sniper bots." GitHub has many of these, and some ask you to paste a wallet private key before they will run. A private key is full control of the assets. Do not hand it to a script you cannot read. A high star count does not make that safe.
- A flattering backtest is not a profit. Fees, slippage, and overfitting all make a live account look worse than the backtest.
08 · How to choose, by situation
- You write Python and want a serious strategy: Freqtrade. If research and validation matter more, choose Jesse.
- You do not want to write code, and you only want a grid or a recurring buy: OctoBot.
- You already use OKX and want an assistant to check prices and positions: OKX Agent Trade Kit, read-only first.
- You do not have a strategy yet: do not install a bot. Start with open-source tools by job, from data and a backtest.
More bots are in the trading bots category. Each tool page shows when the project was last updated.
Frequently asked questions
How to choose a trading bot?
If you write Python and want a real strategy, start with Freqtrade. If you care more about research and validation, choose Jesse. If you do not want to write code and only want a grid or a recurring buy, choose OctoBot. If you already use OKX and want an assistant to read prices and positions, use OKX Agent Trade Kit and start in read-only mode. Conversational tools such as CloddsBot can misread an instruction, so that group is the highest risk. If you do not have a strategy yet, do not install a bot. Whichever you pick, start in a simulated mode.
Who it is for?
This guide is for someone who already has buy and sell rules and wants a program to send those orders without watching the screen. Python users can start with Freqtrade or Jesse. People who will not write code, and only want a grid or dollar-cost averaging, should look at OctoBot. People already on OKX who want an assistant to check prices and positions should look at OKX's toolkit, in read-only mode first.
Who it is not for?
This guide is not for someone who has no strategy yet and hopes the bot will make the money. It compares open-source programs that can send an order. It is not the backtesting-framework comparison, not the China broker-terminal comparison, and not the how-to-choose tool order. A bot executes. It does not earn.
What are the three categories of open-source trading bots?
Strategy frameworks you code yourself: Freqtrade and Jesse. Ready-to-configure bots: OctoBot, for grids, dollar-cost averaging, and similar setups. Conversational tools that turn a sentence into an order: CloddsBot and OKX Agent Trade Kit. The third group is the easiest to start and the easiest to misread.
What should I do before using real money?
Run a simulated mode for at least two weeks and compare it with the backtest. Create an API key with trade permission only, not withdrawal permission. Check that the project is still maintained. Do not give a wallet private key to a script you cannot read. A backtest that ignores fees, slippage, or overfitting will look better than a live account.
This is not investment advice. Crypto and leveraged trading can lose the entire stake. Features, licences, and supported exchanges follow each project's own notice.