Hey hunters! 👋 Builder here. We built Wuzo because getting real financial advice in Brazil means paying R$300/hr for a CFP — or getting generic AI responses that don't know your actual numbers. Our approach: 8 specialized agents, not one chatbot. Each has a dedicated system prompt, calibrated temperature and structured output format. The Finance Parser runs at temp 0.05 for maximum JSON precision. The Academy Educator runs at 0.45 for engaging, creative content. Same model — completely different agents. Two technical decisions that made the biggest difference: ① 3-pass generation. LLMs stop early even at 70B params. We split each analysis into 3 focused calls — diagnosis, quantitative analysis, recommendations — and concatenate. Result: 1,000–1,500 words consistently, matching a paid consultant. ② No cache proxies for prices. Previous version used allorigins.win as a CORS proxy — it was serving 3-day-old BTC prices. Replaced with a 3-source cascade: CoinCap → Coinbase → Blockchain.info. All have native CORS. Always fresh. Free tier is genuinely useful — 5 analyses/month covers monthly tracking for most users. Referral gives +3 analyses to you and +2 to your friend, no limit. What's next: open banking integration via Brazil's Open Finance API — no more manual statement uploads. Try it → https://app.wuzo.com.br Would love your feedback on the agent architecture specifically. What would you add as a 9th agent? 🙏

WUZO

The financial AI your bank doesn't want you to have.

The financial AI your bank doesn't want you to have.