StockAI India
Codevally
FinTech / AI
2 Months
2
Overview
Indian retail investors face an information overload problem — technical indicators, quarterly filings, FII/DII flows, and breaking news all move stock prices, but no consumer-facing tool fuses them across NSE and BSE coverage. Most traders end up copying tips from Telegram channels or watching lagging broker terminals, with no visibility into why a stock should be bought or sold.
Codevally built StockAI India, a multi-agent AI platform that turns any NSE/BSE ticker into an institutional-grade BUY, SELL, or HOLD recommendation in around 25 seconds. A LangGraph orchestrator runs four specialist agents in parallel, and a master agent synthesizes their signals into a single weighted decision — with full transparency over the evidence and risk flags behind it.
The Solution delivers institutional-grade equity analysis by providing:
- Technical Agent: Multi-timeframe RSI, MACD, EMA stacks, ADX, OBV, and automatic candlestick pattern detection.
- Fundamental Agent: Sector-relative P/E, P/B, ROE, debt-to-equity, earnings and revenue trends with LLM-generated interpretation.
- Market Context Agent: FII/DII flows, Nifty/Sensex/BankNifty regimes, India VIX, global indices, and sector relative strength.
- News Sentiment Agent: NewsAPI, Economic Times, and Reuters feeds processed through VADER pre-filtering plus LLM event detection.
- Master Agent: Weighted synthesis of all four agents into a conviction-scored BUY/SELL call with target price, stop loss, key evidence, and risk flags.
This Codevally product condenses what would take a research analyst hours of cross-referencing into a single 25-second pipeline — every recommendation ships with explicit evidence and risk flags, so users always see why the model called BUY or SELL, no black-box scores. The fuzzy ticker resolver, free public access, and Next.js + Tailwind interface remove the last barriers between retail investors and institutional-grade decision support.
Bringing institutional-grade equity analysis to every Indian retail trader.
01. Discovery & Architecture
Mapped Indian retail trader pain points, defined the four-agent decomposition, and selected LangGraph for parallel orchestration with fall-through guarantees.02. Data & Agent Engineering
Wired yFinance, NSE/BSE feeds, NewsAPI, and Economic Times RSS into typed agent contracts. Built indicator stacks, VADER pre-filtering, and LLM prompt chains per agent.03. Master Synthesis & UX
Designed the master agent's weighted scoring, conviction model, and risk flag schema. Built a Next.js interface that surfaces evidence first, score second.04. Performance & Launch
Optimized parallel agent latency to ~25 seconds, added fuzzy ticker resolution, and shipped a public-facing analyzer free for NSE/BSE users.












