OUR WORK
AI trading systems built on real capital, not slide decks.
Every system here solves a problem we hit ourselves — then hardened to production grade. We ship infrastructure that trades, remembers, and protects.
S
Sean Peng
CEO, Mnemox AI · Taipei
1,510
Tests Passing
4
Shipped Systems
35
Countries Reached
780+
GitHub Stars
THE PROBLEM
AI trading agents forget everything between sessions
Every MCP tool teaches AI how to execute trades. None teach AI how to remember why. Prop firms spend $100K+/year on compliance tools that weren't built for AI agents.
1,233Tests
17MCP Tools
SHA-256Audit Trail
TradeMemory Protocol
THE PROBLEM
No single screen to command an AI agent fleet
Managing multiple AI agents across trading, research, and ops meant tab-hopping between terminals. No visibility into what agents decided or why.
5Agent Types
Real-timeWSS Streaming
NexusOS
THE PROBLEM
4 gold strategies, 1 account — without blowing up
Running multiple strategies simultaneously on XAUUSD requires portfolio-level risk control that most MT5 EAs don't have. One bad strategy can wipe the others' gains.
9 layersRisk Control
-3%/dayMax Drawdown Cap
NG_Gold
OPEN SOURCE & TOOLS
Idea Reality
Pre-build reality check for AI agents. Scans 5 platforms in parallel.
457 ★
GitHub Stars · 35+ Countries
OPEN SOURCE & TOOLS
Overfitting detection for trading strategies. Published on PyPI.
PyPI
Open Source Package
OPEN SOURCE & TOOLS
Prevents LLM echo chambers in RAG pipelines. Forces negative case exposure.
MIT
Licensed · Research-backed
Have a trading system that needs building?
We scope, build, and ship — from audit trails to full AI war rooms.
▸ Founder-led builds — limited availability
4 shipped systems · 1,510+ testsFounder-led — limited availability
Book a Call