flemosr

bitcoin

Hey. I wrote a tool that proved to be very useful to me, so I made it FOSS

It is called [`quantoxide`](https://github.com/flemosr/quantoxide) and consists of a Rust framework for developing, backtesting, and deploying algorithmic trading strategies for Bitcoin futures on [LN Markets](https://lnmarkets.com/). It is built on top of [`lnm-sdk`](https://crates.io/crates/lnm-sdk), a LN Markets Rust SDK that I released at the same time.

It gives you a complete workflow:

1. develop your trading strategy logic
2. backtest it against historical price data locally (no API keys needed for this)
3. iterate, validate performance
4. deploy it live (you can start locally or just containerize it and deploy to a VPS for 24/7 operation)

Basically, it handles all the bot infrastructure, syncing historical data to your database (identifying and filling gaps automatically), provides TUIs (Terminal User Interfaces) to monitor each stage, and manages live execution. So you can just focus on strategy.

**Backtest TUI**:

![Backtest TUI Demo](https://raw.githubusercontent.com/flemosr/quantoxide/refs/heads/main/assets/v0.1.0_backtest-tui_20s.gif)

**Live Trading TUI**:

![Live TUI Demo](https://raw.githubusercontent.com/flemosr/quantoxide/refs/heads/main/assets/v0.1.0_live-tui_45s.gif)

I tried to make its API and docs "AI friendly", and added a [specialized quickstart prompt](https://github.com/flemosr/quantoxide/tree/main?tab=readme-ov-file#ai-quickstart) to the repo. This prompt gives agents all the context they need to go from 0 to a working demo with a basic strategy in just a few minutes (~7min in my tests).

It also instructs them to create a persistent context for consistency across sessions, and a backtest binary with LLM-friendly output so they can even run backtests autonomously, if you wish. This allows them to iterate according to results, implementing some basic optimizations. It's *quite fun* to see it working.

The prompt should give good results for setup using the mainstream web UIs (Grok, ChatGPT, etc.), but to leverage the automated optimizations I'd recommend going with Claude Code or a similar agentic platform. You should have privacy in mind if considering using these tools to work on a strategy "for real", but they can really improve the experience of getting started and seeing how everything fits together.

I hope you can leverage it to have some fun playing around with some strategies, or to make some sats.

Feedback is welcome, and I wish you a Merry Christmas!

Repository: https://github.com/flemosr/quantoxide