✓ Checked 2026-10-17
Home › AI trading bots › Open source
Open-source AI trading bots (2026)
⚠ Risk note. No trading bot, AI or not, can guarantee returns. Backtests and "AI robot" track records are hypothetical and often overfit; grid and DCA bots can lose heavily in trending or crashing markets; leverage can liquidate an account. Use paper or demo mode first, give exchange API keys trade-only permission (never withdrawals), cap position sizes, and treat any product promising fixed daily or guaranteed profits as a likely scam. Nothing here is financial advice.
| Bot | License | Best for | AI / agent hooks | Repo |
|---|---|---|---|---|
| Freqtrade | GPL-3.0 | Crypto strategy bots, spot and futures | FreqAI (ML, reinforcement learning); REST API, Telegram | GitHub |
| Hummingbot | Apache-2.0 | Market making, arbitrage, 50+ CEX/DEX connectors | Condor AI harness; agentic strategies | GitHub |
| OctoBot | GPL-3.0 | UI-driven DCA, grid, TradingView and AI bots | ChatGPT/AI strategy modes | GitHub |
| Jesse | MIT | Python research and backtesting with live trading | Built-in MCP server for Claude, Cursor and other agents | GitHub |
| Superalgos | Apache-2.0 | Visual bot design plus a social signals network | Visual scripting; signals network | GitHub |
Running an open-source bot with an AI agent
- Start in dry-run or paper mode (Freqtrade
dry_run: true, OctoBot virtual money, Hummingbot paper connectors). - Create exchange API keys with trading only, no withdrawals, and IP-whitelist your server.
- Let the agent propose strategy changes and backtest them; keep a human approval step before anything goes live.
- Put hard limits (max position, daily loss stop) in the bot config, not in the prompt.
Want a backtester instead of a bot? See backtesting platforms like NautilusTrader and QuantConnect LEAN.
✓ Checked 2026-10-17
✓ Checked 2026-10-17
✓ Checked 2026-10-17
✓ Checked 2026-10-17
FAQ
Which open-source trading bot is best?
Freqtrade is the most popular (about 55k GitHub stars) for crypto strategy bots; Hummingbot for market making and DEX/CEX connectors; Jesse for clean Python research with a built-in MCP server; OctoBot if you want a UI and an optional hosted cloud.
Are open-source trading bots safe?
The code is auditable, and your keys stay on your machine, but you are responsible for hosting, updates and risk limits. Run dry-run/paper mode first and use trade-only API keys.
Can I connect an AI agent to an open-source bot?
Yes. Jesse ships a built-in MCP server, Hummingbot has the Condor AI harness, and Freqtrade exposes a REST API an agent can call.