SwarmTrade is an autonomous AI trading machine. The Queen AI runs a continuous decision loop — scanning market conditions, reasoning over candidate setups, and placing trades — all without manual intervention. Every decision is logged, every position is risk-bounded. Depending on your account configuration, the Queen can trade equities, options, or both.
Platform Architecture
Queen Chat — Natural language interface — command the platform, run research
Queen AI — Autonomous trading engine — observe, filter, decide, act, learn each cycle. Trades equities, options, or both
Scout Bees — Real-time intelligence agents: news, sentiment, technical, opportunity (+ options-flow when options are enabled)
Quant Pre-filter — Scores every options candidate for probability of profit, EV, and liquidity before the AI sees it (options mode)
Risk Controls — Hard position limits, spread width gates, daily loss caps, naked-short blocks
The Queen AI
The autonomous intelligence at the center of SwarmTrade. She observes, reasons, acts, and learns.
1
Observe
The Queen reads the full account state: open positions with live Greeks, buying power, daily P&L, and any active position alerts (DTE warnings, profit targets hit, stop thresholds reached). She also consumes fresh scout findings and her memory of past patterns.
2
Scan & Pre-filter
For options trading, the quant engine fetches the live options chain for each watchlist symbol and scores every plausible defined-risk spread. Only candidates that pass all hard gates — positive expected value, sufficient probability of profit, adequate liquidity, minimum spread width, and position sizing within account limits — are surfaced to the AI. For equity trading, the Queen evaluates direct buy/sell candidates based on scout signals and market regime.
3
Decide (via AI reasoning model)
The Queen sends a structured prompt to a reasoning AI model containing the pre-filtered candidate table, current positions, market regime, IV conditions, scout intelligence, and her own trading rules. The model returns a structured JSON decision: open a specific spread from the candidate table, close an existing position, or observe and wait.
4
Act
Approved entries are submitted as multi-leg MLEG combo orders with a net debit or credit limit. Closes use the same structure. A pending-order steward monitors every live order and cancels anything unfilled within the cycle window to prevent stale exposure.
5
Log & adapt
Every cycle is written to the Activity feed with the full reasoning behind each decision — what she opened, closed, or skipped and why. Open positions carry a running thesis the Queen re-evaluates against fresh conditions on each subsequent cycle.
Cycle Frequency
The Queen runs cycles on a configurable interval (default: ~30 min during market hours). Each cycle is fully logged in the Activity feed.
Default Action: Observe
The Queen's default is patience. She opens a position only when all entry gates are satisfied simultaneously. Most cycles result in 'observe'.
Swarm Trading
How the Queen selects, enters, manages, and closes positions — and the rules that govern every decision.
Options Entry Gates — all must pass before any position opens
When trading options, the Queen can only open a new position when every gate below is satisfied simultaneously. If any gate fails, the cycle returns observe and no trade is placed.
Scout signal — At least one scout finding with high urgency and directional conviction must be present for the target symbol.
Positive expected value (EV) — The quant pre-filter must compute a positive EV for the candidate spread — probability of profit × max gain minus probability of loss × max loss > 0.
Probability of profit ≥ 55% — Every candidate is scored using Black-Scholes risk-neutral probability. Spreads below the POP threshold are discarded before the AI sees them.
Adequate liquidity — Both legs of the spread must have a bid-ask width of less than 15% of the mid price. Illiquid options are rejected — wide spreads mean excessive slippage on entry and exit.
Minimum spread width — The strike spread must be wide enough (at least $5 or 1% of the underlying price, whichever is larger) to avoid the time-value imbalance that can cause a correctly-directional trade to show a mark-to-market loss.
IV regime alignment — Implied volatility rank determines structure: low IV favors buying premium (debit spreads); high IV favors selling premium (credit spreads). Strategies that contradict the current IV regime are rejected.
Market regime: bullish or neutral — The Queen checks market breadth — advance/decline ratio, percentage of assets above key moving averages, and VIX level. In a confirmed bearish regime, long-premium bullish entries are blocked.
Position size within limits — The max loss on any single spread cannot exceed 15% of account equity. The total premium at risk across all open positions is capped at 75% of equity.
Same-cycle close rule — If the Queen closes any position at a loss during a cycle, she cannot open a new position in that same cycle. Rolling a loser straight into the next trade is loss-chasing, and it compounds drawdowns on small accounts.
What the Queen trades
The Queen's instrument universe is determined by your account settings. In equity mode she trades stocks and ETFs directly. In options mode she uses defined-risk vertical spreads by default — two-leg structures where the maximum possible loss is fixed at entry — with single-leg long calls or puts available for high-conviction directional setups. Both modes can be enabled simultaneously. Naked short options (undefined downside risk) are never used.
Bull call spread
Buy a lower-strike call, sell a higher-strike call. Profits if the underlying rises above the long strike by expiry. Cost is the net debit; max loss is capped at that debit.
Bull put spread
Sell a higher-strike put, buy a lower-strike put. Collects premium upfront; profits if the underlying stays above the short put strike. Max loss is the spread width minus premium collected.
Bear put spread
Buy a higher-strike put, sell a lower-strike put. Profits if the underlying falls below the long put strike. Used when the Queen has a bearish thesis with confirmed regime support.
Bear call spread
Sell a lower-strike call, buy a higher-strike call. Collects premium; profits if the underlying stays below the short call strike. Used in high-IV, range-bound environments.
Position management & exits
Profit target: +50% — When a position reaches +50% of its maximum possible profit, the Queen takes profits. Capturing half of max profit at the midpoint of a trade's life has a strong risk-adjusted basis.
Stop loss: −50% of debit / −200% of credit — Long premium (debit spreads) are cut when the net spread value drops to 50% of the entry debit. Credit spreads are cut if the loss reaches 2× the premium originally collected.
DTE-based management — Positions approaching 7 days to expiry enter a managed-exit window. At 0 DTE, all open positions are closed before market close to avoid assignment risk.
ITM spread exception — If a bull call spread has both legs in-the-money (underlying above the short strike), the position will converge to its full maximum value at expiry. The Queen evaluates each cycle whether the underlying is still above the short strike before deciding to hold or cut.
Regime-break close — If the market regime shifts against the thesis of an open position (e.g. a bullish spread opened during a bull regime, and breadth deteriorates sharply), the Queen may close early to avoid holding through a confirmed turn.
Max loss per trade
No single spread can risk more than 15% of account equity. Sized automatically by the quant pre-filter before the AI makes a decision.
Daily loss cap
If realized losses on the day exceed a configurable threshold, new entries are blocked for the remainder of the session. Open positions are not force-closed.
RTH-only MLEG orders
Multi-leg combo orders are only submitted during Regular Trading Hours (9:30 AM – 4:00 PM ET). Pre-market and after-hours submissions are blocked — Alpaca rejects them.
Paper trading by default. SwarmTrade runs on Alpaca paper trading (simulated capital) until you connect a live account. Use Settings → Broker to add live Alpaca credentials. Validate Queen performance in paper mode before enabling live trading.
Options Trading
When options are enabled, the Queen adds a defined-risk options book on top of (or instead of) her equity trading.
Options are opt-in. By default the Queen trades equities only. You enable options in the Queen tab (see Configuring the Queen), where three switches shape how she uses them:
Allow Options
The Queen may trade both equities and options, choosing the right instrument per setup. This is the master switch — turning it on activates the options chain pipeline, the quant pre-filter, and the options-flow scout.
Options Only
Restricts the Queen exclusively to options — she will not open equity positions. Use this when you want a pure options book.
Trade Options Spreads
When on, the Queen prefers defined-risk vertical spreads (two-leg, capped loss). When off, she uses single-leg long calls/puts for directional conviction.
In options mode every candidate passes through the quant pre-filter before the AI ever sees it — probability of profit, expected value, liquidity, spread width, and IV-regime fit are all scored first (see the entry gates above). The Queen never writes naked short options; downside is always defined.
Options data. For richer chain snapshots you can add an options-data API key in the Queen tab. Without it the Queen falls back to the broker's chain data, which is sufficient for spread construction but less detailed for flow analysis.
Covered Calls (income overlay)
Selling calls against shares you already hold to harvest premium — with hard risk guards for the days it goes against you.
A covered call is selling a call option against ≥100 shares you already own of the same underlying. You collect the premium up front. If the stock stays below the strike, the call expires worthless and you keep the premium — pure income. Because the shares back the call, it is never a naked/unlimited-risk short: the worst case is your shares get called away at the strike (a locked-in gain when the strike is above your cost basis).
How the Queen runs the strategy
1
Find coverage
Each cycle she lists your holdings with ≥100 shares — those are eligible. She sells at most 1 call per 100 shares, never over-writing your coverage.
2
Sell a high-probability call
She picks an out-of-the-money strike with a high probability of expiring worthless — the further OTM (and the lower the delta), the safer. This is the core edge: most days the call decays to zero and you keep the premium.
3
Harvest or manage
If the call decays to near-worthless she buys it back cheap to lock the win and free the shares. If it moves against you, she manages it against a hard stop (below).
4
0-DTE is a valid mode
A popular variant is selling same-day (0-DTE) calls every session for daily income. That's supported — the risk isn't the 0-DTE itself, it's the occasional 'bust day' when the underlying rips through the strike, which the risk guards are built to contain.
Risk guards (hard-enforced, not just guidance)
The premium you collect is capped, but a call that goes deep in-the-money can cost far more than that to buy back. These guards contain that tail:
Coverage requirement — A covered call is only allowed when you hold ≥100 shares per contract. Naked/uncovered short calls are always blocked.
Min OTM % (probability floor) — The strike must be at least this far above spot. Further OTM = higher probability the call expires worthless. Blocks selling at-the-money or in-the-money calls where a small move blows through the strike.
Buyback stop (the bust-day exit) — A hard, mechanical stop: when the short call's loss reaches a set multiple of the credit you collected (default 2× = −200%), the position is flagged STOP and the Queen buys it back — capping the loss instead of riding it to a deep-ITM close.
Min DTE (optional) — An optional floor on days-to-expiry. Left at 0 so daily 0-DTE income works; raise it only if you want to forbid same-day expiries entirely.
What you gain
Steady premium income from theta decay, harvested on the many days the call expires worthless. In a choppy or range-bound tape this is a reliable overlay on shares you're already holding.
What you give up
Upside above the strike — if the stock rallies hard, your shares are capped at the strike and the short call must be bought back at a loss. The buyback stop caps how large that loss can get.
Where to configure. Enable covered calls and tune the guards — Min OTM %, Buyback stop, and Min DTE — in the Queen tab → Instrument Mode. You can also steer the strategy with a plain-English mandate (e.g. “run a daily 0-DTE covered-call income strategy on SPY”).
Configuring the Queen
Everything you can tune from the Queen tab on the Swarm page.
The Queen tab on the Swarm page is the single home for every setting — the AI model and your keys, timing, risk, instruments, your watchlist, and a plain-English mandate. Changes apply on the next cycle. (Connecting your broker and going live is handled separately in the Connect & Deploy dialog.) The main groups:
Instrument mode
Allow Options — Trade equities + options. Master switch for the whole options pipeline.
Options Only — Restrict to options exclusively (no equity positions).
Trade Options Spreads — Prefer defined-risk vertical spreads vs. single-leg long calls/puts.
Timing
Queen Cycle — How often the Queen runs a full observe→decide cycle (default ~30 min during market hours).
Scout Cycle — How often the scout bees refresh their intelligence (default 30 min).
Scouts Enabled — Turn the entire scout intelligence layer on or off.
Trading Horizon — Intraday, swing, or long-term — biases how the Queen frames setups and how long she holds.
Risk & stops
Stop-Loss Mode — Let the Queen decide, use a trailing stop, or a hard stop.
Stop Value & Unit — The stop level, expressed as a percent or a dollar amount.
Position Sizing — How aggressively the Queen sizes into a setup, within the hard per-trade and total-risk caps.
Covered Calls — Optionally let the Queen sell short calls against shares she already holds (≥100) as an income overlay.
Focus & mandate
Watchlist — The symbols the Queen considers. An optional watchlist-only toggle restricts her strictly to these names.
Mandate — Plain-English hard directives she must obey every cycle (e.g. "only trade SPY" or "buy AMD under 500").
AI Model & Keys — Pick the brain — Gemini, OpenAI, Anthropic, or Grok — and bring your own API keys.
Models & keys. The Queen runs on Gemini by default, and you can switch her brain to OpenAI, Anthropic, or xAI Grok. Supply your own model API key and an optional options-data API key in the Queen tab — bring-your-own-keys keeps your usage on your own quota.
Scout Bees
Autonomous intelligence agents running continuously, feeding real-time market data to the Queen.
Options-Flow Scout (options mode only). When the Queen is allowed to trade options, a fifth scout activates and surfaces unusual options activity — sweeps, large block trades, and implied-volatility signals — as a corroborating signal for her options setups. In equity-only mode this scout is automatically disabled so options framing never leaks into the Queen's reasoning.
Scout findings are visible in the Swarm page → Scouts tab. Fresh findings feed into the Queen's next decision cycle automatically.
Deep Scan
Ask the Queen to run a full multi-agent stock research report on any ticker.
Deep Scan is SwarmTrade's on-demand research engine. The Queen orchestrates four specialist agents in parallel, each using Gemini with live Google Search grounding to retrieve real numbers — not hallucinated estimates.
The Queen generates bespoke mission briefs for each of the 4 agents, tailored to the specific ticker and its known characteristics.
2
Execute (parallel)
All 4 agents run simultaneously, each calling Gemini with Google Search grounding. Each returns a structured JSON report with stance, confidence, key signal, real metrics, and source URLs.
3
Synthesize
The Queen folds the 4 reports into a final verdict: Strong Buy / Buy / Hold / Sell / Strong Sell, with risk level, potential upside, and a target price note.
Try it now: Open Queen Chat and type Run a Deep Scan on AAPL — the Queen will dispatch the swarm and return a full research card in ~60–120 seconds.
Queen Chat
A full natural-language interface to command the entire platform.
The Chat page is the primary way to interact with SwarmTrade. The Queen understands your intent, calls the appropriate tool behind the scenes, and responds in natural language with structured data cards.
How is my portfolio doing?Shows running status, active strategies, P&L, win rate, and equity/cash.
What strategies are running right now?Lists all active strategies with fitness scores, P&L, and status.
Run a Deep Scan on NVDADispatches the 4-agent research swarm and returns a full analysis card.
Any fresh scout findings?Shows the latest intelligence from all 4 scout bees.
Start the swarmActivates the Queen's autonomous trading loop.
Stop the swarmGracefully halts the Queen's cycle (no positions are force-closed).
Show recent activityLists the Queen's recent decisions and actions with timestamps.
Queen Voice
Talk to the Queen out loud, and let her speak her answers back in her own voice.
On the Swarm page a floating voice orb lets you have a hands-free conversation with the Queen. Tap it and talk — she transcribes what you say, answers, and speaks the reply aloud. She can also narrate notable swarm events as they happen (a new position, a close, a panic alert) so you can keep an ear on the hive while you work.
Talk to her (speech-to-text)
Tap the orb to start listening; it streams your words live, then sends them to the Queen when you tap again to send. Uses your browser's built-in speech recognition — no extra setup, but it needs a supported browser (Chrome/Edge work best) and microphone permission.
She speaks back (Eve TTS)
The Queen answers in Eve's voice through your xAI Grok key. Her spoken replies and event narration route through a secure server-side proxy so your key is never exposed to the browser.
Setting it up
1
Add your xAI Grok key
Voice is gated on your own Grok key (bring-your-own-keys). Add it in the Queen tab on the Swarm page. The voice orb only appears once a valid Grok key is connected AND your browser supports speech input.
2
Allow microphone access
The first time you tap the orb, your browser asks for microphone permission. Grant it so she can hear you. Chrome and Edge have the best speech-input support; if your browser doesn't support it, the orb won't appear.
3
Tap to talk
Tap once to start listening, speak your request, then tap again to send. She thinks, replies in text, and speaks the answer. Tap again while she's speaking to interrupt.
Anything you can type, you can say. The orb runs your speech through the same Queen chat brain as the Chat page, so voice commands like "how is my portfolio doing?", "run a deep scan on NVDA", or "stop the swarm" all work out loud.
Portfolio & Risk
How SwarmTrade protects capital and manages exposure.
Defined-risk only
Every position entered is a spread — max loss is fixed at the net debit (or spread width minus credit). The account can never lose more than the premium paid on any single trade.
Per-trade size cap
The quant engine uses a fractional Kelly criterion to size each trade. Max loss on any one spread is hard-capped at 15% of account equity regardless of Kelly output.
Daily loss circuit breaker
Once realized daily losses hit the configured threshold, new entries are blocked for the rest of the session. Existing positions run to their own stops.
Pending order steward
Every submitted order is monitored. Orders that remain unfilled past ~25 minutes are automatically canceled — no stale limit orders left open overnight.
Paper trading first. The best way to validate the Queen's behavior is to observe several weeks of paper cycles. Check the Activity feed, the Hive Calendar, and the Performance tab to understand her decision patterns before committing real capital.
Manual Terminal
Trade your own account by hand — stocks, options, and crypto — alongside the swarm.
The Terminal is a full trading cockpit: a live chart with drag-to-trade order lines, an options chain with a payoff diagram, a tape, positions and orders, and an AI copilot that can see the same chart you do. Orders route to the same Alpaca account as the swarm, and are tagged so your trades stay out of the Queen's performance stats.
Hands off my trades
On by default. Positions you open in the Terminal are hidden from the Queen entirely — she cannot stop, close, or manage them. Turn it off in Queen config only if you want her to protect your manual trades too.
Everything is staged first
Chart drags, the command bar, and the AI copilot all fill the order ticket rather than sending. Every route to the broker passes the same confirm step.
Only trades placed in the Terminal are protected
“Hands off my trades” works by tagging each fill as manual when it is placed through this platform. An order you place directly in the Alpaca app or website has no such record, so the Queen sees an ordinary position in her account, treats it as hers to manage, and may attach a stop to it or close it. If you want a position left alone, open it from the Terminal — or stop the swarm before trading elsewhere.
Broker Account
Connecting Alpaca, and switching between accounts.
SwarmTrade connects to Alpaca with an API key pair, stored encrypted against your account. The swarm and the Terminal both use it, so whichever account those keys belong to is the account everything trades in.
1
Open Settings → Alpaca
The Alpaca card holds your API key, secret, and a Paper / Live selector.
2
Paste the key pair for the account you want
Generate it from that specific account in the Alpaca dashboard. Stored keys show as **** — type over them to replace.
3
Press Test Connection
It reports the account NUMBER (e.g. PA3XRGP2BD4W), equity, cash, and buying power for the keys entered. Check that number matches the account you intended before saving — an API key id starts with PK and is not the account number.
4
Save
A running swarm reconnects to the new account immediately — no restart needed. If the swarm is stopped, it uses the new keys the next time you start it.
Switching accounts means swapping keys, not picking from a list
An Alpaca key pair is scoped to exactly one account, so there is no account picker — the keys you paste decide the account. If you have two paper accounts, each has its own key pair. Test Connection is the way to confirm you landed on the right one, since it echoes back the account number. Note these are different identifiers: a paper API key id looks like PKOJJPEY… while the account number looks like PA3XRGP2BD4W.
Switching moves the whole swarm
The Queen reads the same credentials, so changing accounts moves her too — new positions, new history, new equity curve. Positions in the old account stay there and she will no longer see or manage them, and your performance figures and leaderboard entry reflect the new account. Stop the swarm before switching so she is not mid-cycle when the account changes underneath her.
Frequently Asked Questions
Still have questions?
Ask the Queen directly — she can explain anything about the platform.