WukongTrade: a production multi-agent AI platform, built by one founder with AI coding agents
WukongTrade is a bilingual platform where a team of named AI agents analyzes markets. Hubert built it with parallel AI coding sessions, the same process we use for client builds.
Bilingual platform
The whole product runs in Vietnamese and English.
A team of AI agents
Named agents (Wukong, Bajie, Sanzang and Wujing) analyze the market and vote. They run on several model providers: OpenAI, Claude, Gemini and Kimi.
Zero LLM Math
The AI never calculates order sizes or stop and target levels. Plain code does that, so the numbers are checked and repeatable.
AI-written daily blog
Daily analysis posts written by AI, with automatic citation checks.
Video pipeline
Claude writes the script, a Vietnamese voiceover reads it and the system renders vertical video.
Dashboards and payments
Dashboards for admins, customers and partners, a partner revenue-share system and Stripe payments.
Docs and stack
An 89-page documentation site. FastAPI, Postgres and Next.js, in Docker on a VPS.
Hubert ran several Claude Code sessions in parallel, each on its own task, following one process: spec, plan, code, deploy.
That is the same process Versarix uses for client builds: a written scope first, then small steps.
Commits in 5.5 months
About 8,200
SourceWukongTrade git history, May to October 2026, read 20 Oct 2026
Automated tests, all passing
20,350
SourceWukongTrade test suite, read 20 Oct 2026
Tasks completed
380+
SourceWukongTrade task tracker, read 20 Oct 2026
Technical documents
200+
SourceWukongTrade docs folder, read 20 Oct 2026
Start with a free scan. We will name 5 places AI could help.