Data-Driven Trader

You have conviction-driven trades. You need research and alerts at 3am when conditions trigger. You do NOT need an autonomous trader making decisions for you.

Atlas is a market research copilot, not an execution engine. It watches the tape, remembers your theses, and tells you when something matters. You place every order yourself.

Why this template ships without execution

Templates set defaults. Defaulting to "executes trades" has too many compliance failure modes:

• Regulated brokers (different KYC per jurisdiction)
• Account loss exposure if a bad signal fires
• Wash trading / pattern day trader rules
• Tax-lot accounting that the agent can't reason about

If you want execution, the architect can add alpaca_trade or binance_trade actions — those plugins exist in the runtime. You just accept the compliance posture explicitly, after the architect walks through the risks.

For most retail traders, research + alerts is the actual value. The trade itself is 30 seconds; the research and timing is the hard part.

What it does

Atlas polls market data every 5 minutes (configurable). It computes the technical indicators you care about (MACD, RSI, VIX, volume regime) and checks them against your defined setups. When a setup fires, it pings you via Telegram with:

• The setup name + which conditions matched
• Current price, volume, ATR
• Your prior thesis on this ticker (from memory)
• Suggested position size given your risk rules
• Honest scoring: "this matches your criteria 4/5 — the VIX context is borderline"

It never says "buy this." It says "the conditions you defined are now true. Your move."

Why this agent shape

• Continuous operation — Atlas polls while you sleep. 3am breakouts get flagged before market open.
• Tools — calls market data APIs, fetches news headlines that may explain a move, computes indicators in-process.
• Persistent memory — remembers your past trades and theses. If you said "I'll cut $NVDA at $X" three weeks ago, Atlas knows.

What you'll configure

After picking this template, the architect will ask you to:

1. Pick a market data source. Polygon (free tier 5 calls/min) is the default. Alpha Vantage and Yahoo Finance also supported.
2. List your watchlist tickers. ~5-20 is typical; more dilutes the signal quality.
3. Define your setups. Plain English: "breakout above 20-day high with volume > 1.5x average and VIX < 25." The architect translates these into evaluable conditions.
4. Set risk rules. Max position size, max daily loss, max concentration per ticker.

What it looks like running

[Telegram, 9:42am]
Atlas — setup fired on $TSLA.

Setup: "20-day breakout + volume confirmation"
- Price: $244.18 (broke $241.50 resistance at 9:31)
- Volume: 2.4x 20-day avg
- RSI(14): 67 (not yet overbought)
- VIX: 14.2 (low-vol regime ✓)
- Your last note on TSLA (March 4): "wait for breakout above 240 with vol;
  size at 5% portfolio max"

Score: 5/5 — full setup match.
Suggested size: 5% portfolio ≈ $XX given current portfolio.
Stop-loss per your rule: $234.80 (4% below entry).

Reply /skip to ignore, /confirm to log as taken (won't execute).

When this template isn't right

• You want fully autonomous trading — Atlas refuses. Add broker actions via the architect, accept the risk yourself.
• You're a long-term holder — the polling cadence is overkill. Use Research Synthesizer with finance feeds instead.
• You trade options or futures with complex Greeks — Atlas handles equities + crypto well; derivatives need custom monitor types.

What you can ask the architect to change

• "Add Alpaca paper trading" — extends to paper-mode execution
• "Watch crypto markets too — Binance feed" — adds the crypto data source
• "Discord instead of Telegram" — swap notification channel
• "Use a local model for analysis (privacy)" — air-gap mode; the market data still has to come from somewhere external