GitHub - hopit-ai/india-trade-cli: Vibe Trading — agentic AI trading platform for Indian markets. 7 LLM analyst agents analyze stocks in parallel, debate bull vs bear, and deliver trade plans. Open-source agentic quantitative researcher for NSE/BSE/NFO.

Agentic trading platform for Indian markets -- 7 AI analyst agents analyze stocks in parallel, debate bull vs bear, and deliver trade plans with entry, stop, and targets. Open-source LLM trader for NSE, BSE, and NFO.

An agentic quantitative researcher that runs multi-agent AI analysis, backtests quant strategies, places live orders via Indian brokers, and pushes Telegram alerts -- all from a terminal CLI, macOS app, or web interface.

7 AI analyst agents run in parallel, debate bull vs bear, and deliver a trade plan with entry, stop, and targets.

More screenshots



Bull/Bear debate with facilitator	Gamma Exposure with flip point and action chips

Live quotes from Zerodha/Fyers	India-specific market patterns

Options scanner with OI analysis

What is Vibe Trading?

Vibe Trading is an agentic AI trading system built for Indian stock markets (NSE/BSE/NFO). It uses multiple LLM-powered agents working together -- a technical analyst, fundamental analyst, options analyst, news/macro analyst, sentiment analyst, sector rotation analyst, and risk manager -- to produce institutional-grade research from a single command.

Key differentiators:

Agentic architecture -- not a chatbot wrapping an API. Seven autonomous agents gather data, analyze independently, then debate and synthesize
Multi-agent debate -- bull and bear researchers argue in multiple rounds before a fund manager synthesizes the final verdict
LLM-agnostic -- works with Gemini (free), Claude, OpenAI, Ollama (local), or any OpenAI-compatible API
Live trading -- connects to Indian brokers (Fyers, Zerodha) for real order placement, not just analysis
Open source -- MIT licensed, fully transparent, no black-box signals

How it works

analyze RELIANCE
        |
  7 AI Analyst Agents  (parallel, agentic)
  Technical . Fundamental . Options . News . Sentiment . Sector . Risk
        |
  Weighted Scorecard + Conflict Detection
        |
  Multi-Round Debate  (5 LLM calls)
  Bull -> Bear -> Rebuttal -> Rebuttal -> Facilitator
        |
  Fund Manager Synthesis  (1 LLM call)
  Verdict: BUY / SELL / HOLD + confidence + rationale
        |
  3 Risk-Profiled Trade Plans
  Aggressive . Neutral . Conservative -- entry, stop, targets, sizing

Standard analyze: 8 LLM calls. deep-analyze: 11 calls, every analyst fully AI-powered.

Installation

Prerequisites

Python 3.11+ (download)
Node.js 18+ (download) -- only needed for the macOS app
Git (download)

Step 1: Clone and install

git clone https://github.com/hopit-ai/india-trade-cli.git
cd india-trade-cli
python -m venv .venv && source .venv/bin/activate
pip install -e .

Step 2: Configure AI provider

Copy the example env file and add your API key:

Edit .env and set your AI provider (pick one):

# Free option -- Google Gemini
AI_PROVIDER=gemini
GEMINI_API_KEY=your_key_from_aistudio.google.com

# Or -- Claude (with Pro/Max subscription, no API key needed)
AI_PROVIDER=claude_subscription

# Or -- OpenAI
AI_PROVIDER=openai
OPENAI_API_KEY=your_key

# Or -- Ollama (free, local, no key needed)
AI_PROVIDER=ollama

Step 3: Run

Terminal CLI (works immediately, no broker needed):

Inside the REPL:

> analyze RELIANCE
> morning-brief
> strategy library

macOS app (requires Node.js):

cd macos-app
npm install
npm run dev

The Electron app auto-starts a FastAPI sidecar on port 8765. Includes streaming analysis, 25+ card types, broker panel, and sidebar quick commands.

Step 4: Connect a broker (optional)

For live market data and trading, connect a broker. See Broker setup below.

Without a broker, the platform uses yfinance (15-min delayed data, no options chain).

AI provider

Provider	Cost	How
Gemini	Free tier	Key from aistudio.google.com
Claude (Pro/Max sub)	Free with subscription	`npm i -g @anthropic-ai/claude-code` -> `claude login`
Claude API	Pay per token	Key from console.anthropic.com
OpenAI	Pay per token	Key from platform.openai.com
Ollama	Free, local	`ollama pull llama3.1`
OpenRouter / Groq / etc.	Varies	Set `OPENAI_BASE_URL` in `.env`

Switch anytime: provider gemini or provider claude_subscription.

Broker setup

Recommended: Fyers (data) + Zerodha (execution)

Function	Broker	Cost
Live quotes, options chain, GEX, IV Smile, historical data	Fyers	Free
Order placement, holdings, positions, funds	Zerodha (Personal plan)	Free
Total		Free

Fyers provides the best free market data API for Indian markets (live quotes, options chain with IV/Greeks, WebSocket streaming). Zerodha handles order execution. Both connected simultaneously.

For full Zerodha API features (live quotes via WebSocket, historical candles), use the Connect plan at Rs 500/month.

Option A: Free setup (Rs 0/month)

Fyers (data): Create free account at fyers.in -> myapi.fyers.in -> Create App (redirect: http://127.0.0.1:8765/fyers/callback)
Zerodha (execution): Create app at developers.kite.trade -> Personal plan (free) -> redirect: http://localhost:8765/zerodha/callback

Add to .env:

FYERS_APP_ID=XXXX-100
FYERS_SECRET_KEY=your_secret
KITE_API_KEY=your_key
KITE_API_SECRET=your_secret

Option B: Fyers only (Rs 0/month)

Use Fyers for everything (data + paper trading). No live order execution.

FYERS_APP_ID=XXXX-100
FYERS_SECRET_KEY=your_secret

Option C: Demo mode (Rs 0, no account needed)

trade --no-broker   # uses yfinance (15-min delayed), no options data

SEBI static IP requirement

Since April 2026, all API orders must be placed from a registered static IP. Check your IP with curl api.ipify.org and register it on your broker's developer console. The platform forces IPv4 connections to avoid IPv6 mismatch issues.

What you can do

Full command reference ->

AI-powered analysis

analyze RELIANCE -- 7 AI analyst agents, bull-bear debate, three trade plans
deep-analyze INFY -- every agent uses AI (11 LLM calls, 3-8 min)
morning-brief -- daily market brief with FII/DII, breadth, news

Live trading

buy YESBANK 1 17 -- quick limit buy (1 share at Rs 17)
sell RELIANCE 5 -- market sell
cancel -- interactive order cancellation
mode live / mode paper -- toggle without restart
trade NIFTY BULLISH -- guided strategy builder with confirmation

Options & derivatives

gex NIFTY -- gamma exposure with flip point and regime
iv-smile NIFTY -- IV skew across strikes
strategy library -- 58 curated strategies (26 options + 32 technical)
strategy use iron_condor NIFTY -- apply template with live data
greeks NIFTY -- portfolio Greeks

Portfolio

holdings -- with today's P&L, overall P&L, and percentages
positions -- open F&O positions
funds -- available margin and utilisation
risk-report -- VaR, volatility, concentration risk

Data & intelligence

flows -- FII/DII flows with directional signals
deals RELIANCE -- bulk and block deals
earnings -- upcoming results calendar
backtest RELIANCE rsi -- strategy backtests on NSE history
pairs HDFCBANK ICICIBANK -- pair trading signals
memory -- past analyses, verdicts, win rate

Alerts & notifications

alert NIFTY below 23000 -- price alerts
telegram setup -- Telegram bot for quotes, analysis, alerts on mobile

macOS app features

Dark theme React UI with Tailwind CSS
Streaming analysis -- AI analyst agents appear live as they complete, debate streams in real-time
Mid-stream context injection -- type "focus on AI deals" while analysis runs, and the synthesis adapts
25+ card types -- Quote, Analysis, GEX, IV Smile, Strategy, Holdings, Funds, Orders, Memory, Morning Brief, Alerts, Risk Report, Delta Hedge, and more
Contextual action chips -- data-aware follow-up questions on every card (e.g. "Put skew +26% at 22,100 -- why?")
Broker panel -- connect Fyers/Zerodha via OAuth from the sidebar
Sidebar quick commands -- one-click access to all tools

Configuration

# .env  (see .env.example for all options)
AI_PROVIDER=gemini              # gemini | anthropic | openai | claude_subscription
GEMINI_API_KEY=AIza...

TOTAL_CAPITAL=200000            # INR -- used for position sizing
DEFAULT_RISK_PCT=2              # max risk per trade (%)
TRADING_MODE=PAPER              # PAPER or LIVE

NEWSAPI_KEY=...                 # optional -- improves morning-brief quality
TELEGRAM_BOT_TOKEN=...          # optional

Position sizing auto-adjusts for VIX: 100% normal -> 85% at VIX 15-20 -> 65% at 20-25 -> 50% above 25.

API keys are stored in the OS keychain (macOS Keychain / Linux SecretService / Windows Credential Locker).

Brokers

Broker	Data	Execution	Status
Fyers	Live quotes, options chain, WebSocket	Order placement	Fully supported (free)
Zerodha	Quotes, historical (Rs 500/mo Connect plan)	Order placement	Fully supported -- live orders tested
Mock / Demo	yfinance (delayed)	Paper trades	Fully supported (`trade --no-broker`)
Angel One	Quotes, holdings	Order placement	Implemented, needs testing
Upstox	Quotes, holdings	Order placement	Implemented, needs testing
Groww	--	--	WIP

Multi-broker mode

Connect multiple brokers simultaneously. Use one for data, another for execution:

Data broker:      Fyers (free, best options data)
Execution broker: Zerodha (your existing demat account)

Both stay authenticated. Currently the last-connected broker becomes primary for all operations. Explicit data/execution routing is planned (#129).

OpenClaw HTTP skills

The macOS app auto-starts the API server. For CLI-only:

uvicorn web.api:app --host 127.0.0.1 --port 8765
# or from REPL: web

Exposes 17+ skills as POST /skills/<name>. Discovery: GET /.well-known/openclaw.json.

Includes: quote, analyze, deep_analyze, backtest, flows, morning_brief, chat (session-aware), options_chain, deals, earnings, macro, alerts/*, iv_smile, gex, strategy, and more.

No auth -- keep on 127.0.0.1 or put behind a proxy.

Telegram bot

/quote RELIANCE      -- live price
/analyze INFY        -- full multi-agent analysis (3-4 min)
/deepanalyze NIFTY   -- deep AI analysis (7-10 min)
/brief               -- morning brief
/flows               -- FII/DII data
/help                -- all commands

Setup: create a bot via @BotFather, add TELEGRAM_BOT_TOKEN to .env, run telegram in the REPL.

Architecture

agent/      AI agent layer -- LLM providers, analyst agents, multi-agent debate, synthesis
brokers/    Indian broker integrations -- Fyers, Zerodha, Angel One, Upstox, mock
market/     Market data -- quotes, WebSocket, options chain, news, FII/DII
analysis/   Technical, fundamental, options Greeks, GEX, IV smile, multi-timeframe
engine/     Backtester, 58-strategy library, risk metrics, alerts, trade executor
bot/        Telegram bot
app/        REPL entry point and command handlers
web/        FastAPI -- OAuth + OpenClaw skill endpoints
macos-app/  Electron + React macOS app with streaming UI

Roadmap

Shipped: 7 AI analyst agents . scorecard . multi-round debate . 3 risk personas . trade memory . backtesting . walk-forward . what-if . FII/DII flows . event strategies . options analytics . VaR/CVaR . DCF valuation . model drift . pair trading . Telegram bot . Fyers WebSocket . paper trading . live trading (Zerodha) . PDF export . OpenClaw skills . 58-strategy library . backtest cache . macOS Electron app . streaming analysis . GEX . IV smile . contextual action chips . buy/sell/cancel CLI . mode toggle . mid-stream context injection

In progress: multi-session UI (#110) . dual-broker mode (#129) . portfolio-aware personalisation (#118, #119, #121) . web search for agents (#149)

Planned: app order placement UI (#130) . ML prediction analyst (#145) . ensemble meta-model (#146) . TradingView webhooks . SEBI compliance layer . memory improvements (#122)

See open issues.

Contributing

pip install -e .
pytest                    # no API keys needed, network tests skipped by default
trade --no-broker         # smoke test

See CONTRIBUTING.md. Most wanted: broker integrations (#80), app order placement UI (#130), integration tests (#79).

Troubleshooting

Problem	Fix
`ModuleNotFoundError: kiteconnect`	Broker SDKs are optional. Use Demo mode (`login 0`).
NSE API returns empty	NSE rate-limits scrapers. Wait a minute and retry.
Fyers `invalid app id hash`	App ID / Secret mismatch. `credentials delete FYERS_APP_ID` then re-login. App ID format: `XXXX-100`.
`No active broker session`	Run `login` or `login 0` (demo) first.
AI commands error	Run `credentials setup` to configure your AI provider.
`keyring` errors on Linux	`sudo apt install gnome-keyring` or set credentials via `.env`.
Tests failing locally	`pip install pytest pytest-mock && pytest -m "not network"`
Zerodha IP not allowed	Register your IPv4 at developers.kite.trade. Run `curl api.ipify.org` to check.
Order rejected (circuit limit)	Price outside NSE circuit band. Use a price within the allowed range.

FAQ

What is agentic trading? Agentic trading uses autonomous AI agents that independently gather data, analyze markets, and make trading decisions -- rather than a single LLM answering questions. Vibe Trading runs 7 specialized agents (technical, fundamental, options, news, sentiment, sector, risk) that work in parallel, then debate before producing a final recommendation.

How is this different from a trading chatbot? Trading chatbots take your question and return an answer. Vibe Trading's agents proactively gather data using tools (live quotes, options chains, FII/DII flows, news), analyze it independently, then engage in structured multi-round debate. The architecture mirrors how an institutional research desk operates -- multiple analysts with different specializations feeding into a portfolio manager.

Can I use this for live trading? Yes. Connect Zerodha or Fyers for live order placement on NSE/BSE/NFO. Use mode paper for paper trading and mode live for real orders. Always start with paper trading.

What LLMs does it support? Any LLM: Gemini (free), Claude (subscription or API), OpenAI, Ollama (local models like Llama), or any OpenAI-compatible endpoint (OpenRouter, Groq, Together, etc.).

Is this financial advice? No. Vibe Trading is an open-source research tool. All trading decisions are yours. See disclaimer below.

Disclaimer: For educational and research purposes only. Not financial advice. Trading involves substantial risk of loss. The authors are not responsible for financial losses.

MIT -- see LICENSE.