Hummingbot

Hummingbot is a framework for designing and deploying automated trading strategies, or bots, that run on many centralized and decentralized exchanges. The stated mission is to democratize high-frequency trading by keeping the codebase public and the community of algorithmic traders around it open.

The client is driven by hbot, a command-line interface installed from source with conda or run inside Docker. It starts and stops a bot, authors and tunes configs, and reports trades, PnL, logs and status as compact Markdown with stable exit codes, which makes it scriptable by agents as well as people. The classic full-screen interactive client is still there for those who prefer it.

Strategies are written at whichever level of abstraction suits the job.

• Scripts: single-file Python strategies, the shortest path to a custom bot, with a basic market making example included.
• Controllers: reusable V2 strategies whose configs can be backtested, deployed and tuned live while the bot runs.
• Executors: self-contained blocks that manage order lifecycles for position, DCA, grid, arbitrage, XEMM, TWAP and LP patterns.
• Paper trading: a paper trade connector simulates against live market data, so a first run needs no API keys at all.
• Condor: an AI harness connecting model-driven decisions to deterministic execution through the Hummingbot API.

Exchange API keys are encrypted behind a keystore password that can arrive by environment variable or standard input, so unattended runs stay possible. Users have generated over 34 billion dollars in volume across more than 140 trading venues.