Simulated equity
$10,000Fake money · no brokerage Invite-only agentic investing
Private agentic
portfolio intelligence.
An intelligent portfolio command center for an elite circle of investors. Put autonomous market insight to work inside your rules, with deliberate controls and complete human authority.
Agent-directed intelligencePrivate membershipHuman-defined guardrails
◆ Protected with password + passkey authentication
Simulation dashboard
Good day.
Your agent can research and manage a guarded portfolio using fake money only.
◒
Strategy mix
50 / 30 / 20Passive · Balanced · Aggressive◷
Scheduled reviews
0Enabled check-insSystem readiness
Built to think. Bound to your rules.
Catch now tests the complete decision and risk pipeline against a recoverable simulated portfolio. Live trading and Robinhood remain unavailable.
✓
Identity securedPassword + passkey
✓
Risk engine activeDeterministic validation
✓
Simulation onlyNo brokerage execution
AI research engine
Market analysis
Research current markets and send proposals through deterministic validation. Approved actions can fill only in the simulated portfolio.
Monthly AI Usage Allowance
Covers AI research usage only. It is not a trading balance, buying power, or trading limit.
—
Loading usage…
Ready for researchHOLD and NO ACTION are successful outcomes.
Recent analyses
Last 25 runsFake money · real guardrails
Simulated Investing
Track paper positions, performance, fills, and every deterministic decision. Nothing on this page can reach a brokerage.
Open positions
Fractional shares supportedRecent activity
Auditable decisionsRestart simulation
History is preservedTransparent by design
How it works
The system uses two layers: the AI proposes a trade, then deterministic code decides whether it is allowed. The AI cannot execute a trade by itself.
01
Research layerAI proposes
Current research, portfolio context, thesis, and risks.
→
02
Safety layerCode validates
Trusted data and every deterministic guardrail.
→
03
Simulation layerPaper fill
Fractional shares and fake money. No brokerage.
-
01
Start safely
Start the review
A manual request or scheduled Cloudflare review starts the same workflow. It first confirms:
- The simulation account is active.
- Global, administrator, and user simulation switches are enabled.
- The U.S. market is in its regular session.
- Market data and AI budgets are available.
If any prerequisite fails, it records a no-action result.
-
02
Authoritative context
Build the research context
The model receives:
Simulated cash and total equityOpen positions and cost basesCurrent trusted price snapshotsRealized and unrealized P&LStrategy targets and current exposureEvery active guardrailExcluded industries and ETF-only preferenceFor an account with no prior fills, it enters bootstrap mode and may propose up to the configured maximum initial purchases.
-
03
Research proposal
AI proposes an action
The research model can return:
BUYInvest the specified dollar amount.SELLSell the specified dollar amount.REBALANCEMove a position to a desired post-trade value.HOLDRetain an existing position.NO ACTIONDo nothing.Each proposal contains a ticker, strategy, confidence score, thesis, risks, and suggested dollar value.
The buy or sell thesis is qualitative research rather than a fixed formula. The model considers market conditions, company or ETF research, portfolio exposure, valuation or opportunity, current P&L, and the strategy's purpose.
-
04
Independent prices
Obtain trusted market data
The system independently retrieves the security classification and latest quote from FMP. The model is not allowed to provide:
- Execution prices
- Share quantities
- Fill results
- Guardrail decisions
Missing, stale, or inconsistent provider data causes rejection.
-
05
Deterministic safety
Independently validate the proposal
The deterministic engine checks:
Supported U.S. stock or ETFRegular market sessionQuote no more than five minutes oldUSD and supported exchangeETF-only settingStock market-cap and excluded-industry rulesInitial-purchase or normal daily trade limitUser daily-loss limit and hard 15% backstopMaximum 15% of equity per tradeMaximum 20% projected exposure per tickerMinimum cash reserve30-day turnoverStrategy driftMinimum and maximum trade valuesAvailable cash or sharesMinimum holding period for salesConfidence thresholds90% Passive75% Balanced70% AggressiveAny failed check rejects the proposal and records the reasons.
-
06
Simulated execution
Calculate the simulated fill
For an approved trade, the engine calculates fractional shares using the trusted quote:
BUY PRICEQuote × 1.001SELL PRICEQuote × 0.999COMMISSION$0.00ExampleAn approved $1,000 buy at a $100 quote uses a $100.10 simulated fill price and derives the fractional quantity from that—not from the model.
-
07
Auditable accounting
Update the portfolio
The workflow atomically records:
Trade intentSimulated fillCash movementPosition quantity and cost basisRealized or unrealized P&LAccount-balance historyActivity explanationRetries cannot create a duplicate fill.
The important distinction
AI finds the opportunity. Code controls the action.
The AI determines whether a buy or sell appears desirable, but the risk engine determines whether that proposal is actually safe and executable. HOLD, rejection, and no-action are all normal successful outcomes.
Your profile & security
Account
Personalize your experience and manage the passkeys that protect your account.
Appearance
Theme
Trusted credentials
Your passkeys
Loading passkeys...
Add another device
Generate a setup code
Open Catch Investments on the new device, choose Add device, and enter this one-time code. The code expires after 10 minutes.
One-time setup code
Protected operations
Administration
Approve members and control system-wide availability.
Global trading
Fail-closed control checked before every future order attempt.
Global simulation
Stops all paper fills without changing future live-trading controls.
OpenAI research
Checking encrypted server configuration…