Overview

This is a self-hosted, non-custodial spot trading bot for Binance, with a web dashboard. You run it on your own machine, against your own Binance API keys. The image ships with no keys, sends your credentials nowhere, and defaults to Binance testnet.

It runs one strategy family, but runs it thoroughly: a DCA / Grid ladder, with an optional Hybrid micro-scalp layered on top. Understanding that strategy is most of understanding the bot, so those sections are the heart of these docs — this page is the map.

Current version — 2.0.1

See the release list for version history and notes.

Why this bot

#	Advantage
1	Non-custodial & self-hosted. Runs on your own machine with your own keys. The bot holds no funds and has no withdrawal permission — your keys never leave your .env.
2	Averages the entry down. A DCA ladder of ever-deeper, ever-larger safety orders pulls your average entry down as price falls, instead of a single all-in buy.
3	One take-profit from real fills. The whole position is closed by a single exit, sized to everything held so far and recomputed from your actual fills — it tracks reality, not the original plan.
4	Scalps the wait (Hybrid). While the grid sits below its exit, small partial closes harvest the up-and-down chop and bank profit that pulls the whole-position exit closer.
5	Reaches deeper than Binance's ~18% order cap. Binance rejects a limit buy placed too far below market (its percent-price filter, ~18%). Because the grid keeps only a limited number of live orders — the Active orders setting — near the price and places deeper rungs progressively as price falls, the ladder extends far below that single-order cap over a falling move.
6	Respects exchange filters automatically. Every order is rounded to the pair's tick size, step size, and minimum notional, so orders aren't rejected by the exchange.
7	Testnet-first, with a safety fallback. Ships with no keys and defaults to Binance testnet; real mode without real keys silently falls back to testnet, so you can't trade real funds by accident.
8	Manual per-order control (Expert Mode). Cancel and re-place individual live orders by hand when you want to intervene.
9	Free, open source, and yours to run 24/7. GPLv3, self-hosted — no subscription, no third party between you and the exchange.

What it does, in one loop

A single cycle on one trading pair (e.g. BNBUSDT) works like this:

1. Enter and average down. The bot lays a ladder of buy orders below the current price. As price falls, deeper rungs fill, each one larger than the last, pulling your average entry down. This is the classic DCA / Grid.
2. Close the whole position at once. Against that averaged entry it rests a single take-profit sized to the entire position held so far. It is recomputed from your real fills, so it tracks reality rather than the original plan.
3. (Hybrid) Scalp the wait. With Hybrid on, while the position sits below that close, a small partial close — the micro-scalp — harvests the up-and-down chop. Each fire is banked, and banked profit pulls the whole-position exit closer.
4. Bank, re-arm, repeat. When the grid finally closes in profit, the cycle ends and the next one begins.

Who it is for

This is an operator's tool, not a one-click product. It assumes you understand limit orders, averaging down, and the risk of holding a falling position. In Expert Mode it even lets you cancel and re-place individual live orders by hand.

Trade on testnet first

In real mode the bot places live orders with real funds. Start on testnet, learn how a cycle behaves, and only then consider real keys. See the Disclaimer.

AI-assisted deployment and coding

Using AI is encouraged, especially for setup, deployment, and repetitive operator work. It usually does not break the project by itself — it helps you launch faster and reduces routine friction.

But always review AI-generated changes with your own eyes. New features and non-trivial edits should be tested on Binance testnet first. AI often gets architecture or integration details wrong — not only because the model can make mistakes, but also because humans frequently describe requirements incompletely.

Treat AI as an accelerator, not as a final authority: inspect the code, verify the behavior, and test the risky paths manually.

How to read these docs

• Getting Started — get it running in a few minutes.
• DCA / Grid — the classic strategy and every control on the dashboard, field by field.
• Expert Mode — manual, per-order intervention.
• Hybrid — the micro-scalp: parameters, the summary bar, the grid marks, and the badge legend.
• Operations — persistence, backups, configuration, updates.