bybit-ws — AI-Native Trading Engine

Bollinger Grid + 8 strategies. REST API + MCP. Built for AI agents.

---

What is it?

A production-grade trading monitor for Bybit that runs 24/7, scans for opportunities, enters positions, manages risk, and logs everything — all accessible through an HTTP API designed for AI agents.

Not a bot. An engine. Your AI agent calls /scan, picks signals, calls /enter. The engine handles execution, SL/TP, risk limits, and reporting.

Quick Start (for AI Agents)

git clone https://github.com/poliakarmai/bybit-ws.git
cd bybit-ws
cp config.example.yaml ~/.config/bybit-ws/config.yaml
# Set BYBIT_API_KEY and BYBIT_API_SECRET
python3 -m bybit_ws.main

from bybit_ws_sdk import Monitor
m = Monitor("http://localhost:8766")

signals = m.scan(mode="long", limit=5)
for s in signals["signals"]:
    if s["score"] >= 7.0:
        m.enter(s["symbol"], "Buy", s["qty"])

Full Quickstart →

Strategies

#	Strategy	Leverage	Trigger
1	BB Grid LONG	3x	BB < 25% Daily, scoring ≥ 5.5
2	BB Grid SHORT	3x	BB > 85% Daily, Tier A/B
3	Junk SHORT	3x	Daily pump ≥ 80%, no SL, hard stop
4	SL Re-entry	3x	Ladder after stop-loss (-5%, -10%, -15%)
5	BB Scalp M5 ⚡	10x	BB touch M5 + RSI filter
6	Mean Revert ⚡	10x	BB% < 5% or > 95% Daily
7	Funding Momentum ⚡	10x	Extreme funding + BB + trend
8	DCA	3x	Add to losing positions at lower levels

All strategies with parameters →

Architecture

AI Agent (Claude Code / Codex / Cursor / Hermes)
    │
    ├── REST API (port 8766): scan, enter, close, health
    ├── MCP Server: scan_market, get_positions, get_metrics
    └── Python SDK: Monitor + WebhookHandler classes
    │
bybit-ws (30s cycle, systemd or Docker)
    ├── 8 strategies with scoring + risk checks
    ├── Position sizing: dynamic % of deposit
    ├── Correlation check: block entries >0.8
    ├── X10 safety: daily loss stop, ATR validation
    ├── Auto SL/TP via trading-stop
    └── Dashboard: SVG with winrate, margin, regime

Full architecture →

Risk Management

• Position sizing: dynamic margin = deposit × risk% / positions × score
• Drawdown guard: auto-close all at -15% from peak
• Daily loss stop: halt at -$50/day
• Correlation block: reject entries if ≥2 positions correlated >0.8
• X10 limits: max 3 losing x10 trades → 24h cooldown
• Cascade protection: market-close if price 2x closer to liquidation than SL
• Banned symbols: config-driven permanent ban

Files

File	For
QUICKSTART.md	5-minute setup
DESIGN.md	Full architecture for devs & agents
STRATEGIES.md	All 8 strategies + roadmap
ERRORS.md	Error reference: what + fix
WEBHOOKS.md	Alert payloads & parsing
CHANGELOG.md	Version history
openapi.yaml	OpenAPI 3.0 spec
bybit_ws_sdk.py	Python SDK
config.example.yaml	Full config with comments

Requirements

• Python 3.11+
• Bybit Unified Trading Account with API keys
• Linux server (VPS $5/mo works) or Docker

License

MIT — see LICENSE.

---

Built for AI agents. Ready for yours.