What this dashboard is
Welcome. This page explains, in plain language, what this tool does and what you are looking at — so the rest of the screens stop feeling like a black box.
In one sentence
This is a paper-trading decision-support system for Indian equities (NSE/BSE). It uses AI agents to research a stock and propose a trade, then stops and asks a human to approve before any order is "placed". Nothing touches real money.
Paper mode means every order is simulated against an in-memory broker. There is no live brokerage account wired in. The orange PAPER badge in the top-left is always on — that is by design, and it is your reminder that this is a sandbox.
What problem it solves
A normal trading idea goes through a lot of hidden steps: gathering news, reading sentiment, checking the chart, checking the fundamentals, arguing both sides, sizing the position so one bad trade can't blow up the account, and finally deciding. This system makes every one of those steps visible and reviewable on screen, and forces a human signature before acting.
So when you open a screen and wonder "what's going on?" — the answer is always: the system is somewhere along a 12-step assembly line, and the dashboard is showing you one station of that line.
The 12-stage assembly line
Every run walks through these stages in order. Each stage is tagged by who is in control:
| # | Stage | Who runs it | What it does |
|---|---|---|---|
| 1 | Market & news feeds | Deterministic | Pulls prices, news, macro data for the symbol |
| 2 | Four analyst agents | AI | News, Sentiment, Technical, Fundamental — each writes a note |
| 3 | Memory store | AI | Recalls past theses & outcomes to give the agents context |
| 4 | Strategy approval | Human | Checks the symbol is inside your approved mandate |
| 5 | Bull vs Bear debate | AI | Two agents argue; a manager scores the conviction |
| 6 | Trade thesis | AI | Proposes direction, entry, target, stop, horizon |
| 7 | Risk & sizing | Deterministic | Pure code sizes the position and enforces caps |
| 8 | Human approval gate | Human | You approve, modify, or reject. Nothing proceeds without a signature |
| 9 | Broker execution | Deterministic | Formats and "places" the (paper) order, records a fill |
| 10 | Kill switch | Human | One click flattens everything and blocks new orders |
| 11 | Portfolio monitoring | Deterministic | Live P&L, exposure, alerts |
| 12 | Post-trade review | AI | Critiques the outcome vs the thesis and saves the lesson |
You do not run these by hand one-by-one. You trigger a run (stage 1) and the system flows automatically up to stage 8, where it parks and waits for you. After you approve, it finishes stages 9–12 on its own.
Why it stops in the middle
Stages 7, 9, and 11 — risk sizing, order routing, and P&L — are plain code with no AI in them, ever. The AI is allowed to suggest a trade, but it is never allowed to size it, route it, or bypass you. The hard stop at stage 8 is the whole point of the design: an order cannot leave without a recorded human approval.
That is why the natural rhythm of using this tool is:
- Trigger a run for a symbol.
- Review what the AI came up with on the approval screen.
- Approve / modify / reject.
- Watch the position and P&L in the portfolio.
The next page gets you to that first run.