Voinista
Watch twelve AI analysts argue your next trade
For: Retail traders and individual quants researching US equities, plus analysts and small funds who want a fast adversarial second opinion and an exportable research dossier.
Voinista is a multi-agent research desk for US equities. You pick a ticker and a date; twelve AI agents then work it in three stages. Four analysts (fundamentals, news, sentiment, technicals) each file a sourced report, a bull and a bear argue opposite sides, and a research manager, a trader, a three-person risk committee and a portfolio manager turn the debate into a verdict.
The product is the reasoning, not just the call. Every run ends with BUY, HOLD or SELL plus a confidence level, a short rationale, the complete debate transcript, an Influence Matrix showing which analysts drove the decision, and an execution blueprint with entry zone, stop and target. A fast run takes about five minutes and a deep run about ten.
It is built for retail traders and individual quants who lack a terminal seat or the time to read a 10-K, and for analysts and small funds who want a fast adversarial second opinion and a dossier they can hand on. Its boundary is firm: no order routing, no custody and no personalised advice. Every output is labelled 'Research, not investment advice.'
Voinista is built by Riaan Kleynhans and the engineering team at The Nine, on the open-source TauricResearch/TradingAgents framework.
12
AI agents per run
3
Free runs a month, no card
€0–€99
Monthly plans (Free, Pro, Quant)
8
Artifacts produced per run
What it does
Twelve-agent desk: four analysts, bull and bear researchers, research manager, trader, aggressive, conservative and neutral risk reviewers, and a portfolio manager
Analysis grounded in 10-K and 10-Q filings and company news, with sources attached
War Room view that shows each agent's progress and the debate as it resolves
Executive verdict (BUY, HOLD or SELL) with confidence, three-bullet rationale and execution blueprint (entry zone, stop-loss, target)
Influence Matrix showing which analysts tipped the decision
Alpha Radar discovery board ranked by market-regime strength, plus search across any US ticker
Auditable Markov regime model (20-day return states, transition matrix, persistence odds, Alpha-Decay Clock) provided as context only, never as a recommendation
PDF dossier export (Quant tier), shareable public run links and full run history
How we architected it
- Next.js 14
- React 18
- TypeScript
- Tailwind CSS
- Geist
- Framer Motion
- Three.js
- cmdk
- @react-pdf/renderer
- Supabase (Postgres, Auth, RLS, Realtime)
- Stripe
- Resend
- PostHog
- Vercel Analytics
- Python
- FastAPI
- LangGraph
- TauricResearch/TradingAgents
- financialdatasets.ai
- Vercel
- Fly.io
- 01
Next.js 14 App Router front end (React 18, Tailwind CSS, Geist) on Vercel, with marketing (www) and console (app) split across two origins by middleware
- 02
Separate Python worker running TauricResearch/TradingAgents on FastAPI and LangGraph, hosted on Fly.io and called only server-side with a shared-secret bearer token
- 03
Market data and SEC filings from api.financialdatasets.ai
- 04
Run lifecycle: a run row is created in Postgres, dispatched to the worker with an idempotency key, the UI follows progress through Supabase Realtime with a polling fallback, and the result is stored as a durable snapshot that survives worker restarts
- 05
Supabase Postgres with row-level security on every table; run history is private to its owner and only runs marked public can be read logged-out
- 06
Multi-provider model menu (Anthropic, OpenAI, Google, xAI, DeepSeek) with credits priced by real token cost; failed or cancelled runs cost nothing
- 07
Stripe subscriptions and top-up boosters, with the webhook as the source of truth for tier; conversion-gated referral engine; PDF output via @react-pdf/renderer
- 08
Public routes tested with Playwright and axe-core at zero WCAG 2.2 AA violations
Want something built like this?
Every build runs the same five gates: spec, architect, build, review, ship. It starts with a 48-hour audit.