Pre-TOKEN2049 edition · Claude Singapore Community
In March, OKX did something no major exchange had done: they open-sourced a toolkit that lets an AI agent trade directly on their platform. Spot, futures, options, grid bots, 80+ tools, MIT licensed, running locally on your machine.
Most of the industry still hasn't looked at it.
This session is a straight, practical hour on what exchange-native AI trading actually is, what it can and can't do, and what it means for how you work.
What we'll cover
What OKX actually shipped. The Agent Trade Kit, how it plugs into Claude Code through MCP, and the real scope of what an agent can reach, market data, order execution, position management, automated strategies.
The security model, honestly. Your API keys stay in a local config file. The model never sees them. Signing happens on your machine. Why that design matters, and where the remaining risk actually sits.
How people are using it. Research and market analysis, monitoring, strategy testing in demo mode, and execution. Where each one genuinely helps, and where it falls apart.
The part nobody wants to say out loud. AI makes mistakes, and you own the losses. What guardrails look like in practice, read-only mode, demo profiles, permission-scoped keys, and why you should never skip them.
Where the edge goes. When every desk can run the same agent on the same public data, what's actually left worth having.
Then we build one, live
Last fifteen minutes: a working agent wired to OKX through Claude Code, pulling live market data, running in read-only mode. From nothing to working, on stage. No laptops needed, just watch how little it takes.
And then, Pre-TOKEN Networking
Traders, founders, researchers, and anyone doing crypto work who wants to understand this properly before TOKEN2049 week takes over.
Run of show
4:00, Doors Open and networking4:30, Talk: Trading Agents5:30, Live build: an OKX agent in 15 minutes5:45, Open networking6:30, Close
Details
Tuesday, 29 September 2026 · 4:30–6:30 PM
Nothing in this session is financial advice. Automated trading carries real risk of loss.