ConvictionOS
View source Open Mission Control
Alpaca paper trading Grounded AI · Deterministic execution

Evidence earns capital.

ConvictionOS is the operating system for accountable AI options trading—turning market evidence into bounded decisions, hard risk checks, and receipts you can verify.

Open Mission Control Explore the product

Paper-only by design. No return promises. No real-money execution path.

Mission ControlGood morning, operator.
Paper mode
Agent statusEvidence gatedAwaiting next cycle
Primary marketSPY optionsMulti-horizon mandates
Risk authorityDeterministicModel cannot bypass
Decision pipeline5 gates
EvidenceGrounded
ThesisStructured
RiskEvaluating
ExecutionPending
Latest receiptABSTAIN

Quote freshness gate protected the paper account.

Evidence-bound decision
Paper-onlyAlpaca execution boundary
Provider-agnosticOpenRouter, Gemini, compatible gateways
Fail-closedUncertainty becomes abstention
AuditableDecision and lifecycle receipts

How ConvictionOS works

From evidence to an auditable decision

One controlled path separates model reasoning from trading authority.

01

Ground the thesis

Point-in-time market evidence, volatility context, and quant signals become a structured thesis—not a free-form order.

02

Enforce the mandate

Deterministic gates check liquidity, quote age, exposure, buying power, structure, and lifecycle policy.

03

Prove the outcome

Submission, rejection, abstention, reconciliation, and exit events produce a clear decision receipt.

One platform, two perspectives

Built for individual investors and quant teams

Clear enough to understand at a glance. Deep enough to inspect every assumption.

Individual investors

Stay in control without living in spreadsheets.

  • Plain-language decisions and abstentions
  • Visible maximum loss and mandate limits
  • Paper-first learning without hidden automation
Quant & portfolio teams

Inspect the system behind every decision.

  • Evidence provenance and structured theses
  • Independent risk and broker reconciliation
  • Strategy promotion with measurable gates

Strategy lifecycle

Conviction is earned in stages.

Research never becomes paper capital because a model sounded confident.

  1. 1Drafted

    Thesis, universe, horizon, and risk rules are explicit.

  2. 2Validated

    Inputs, calculations, and refusal behavior pass checks.

  3. 3Backtested

    Evidence is evaluated without rewriting history.

  4. 4Published

    A versioned mandate is approved for observation.

  5. 5Paper running

    Only eligible strategies receive bounded paper capital.

Safety is the product

Paper capital.
Production discipline.

ConvictionOS treats “do nothing” as a valid decision. If market truth, account state, or model output is incomplete, execution stops.

Explore risk controls
01

AI is advisory

The model proposes a bounded thesis. It never owns position sizing or broker authority.

02

Risk is deterministic

Freshness, liquidity, exposure, mandate, and lifecycle checks are explicit and repeatable.

03

Broker state wins

Alpaca paper acknowledgements and reconciliation determine what actually happened.

04

Receipts preserve context

Decisions retain the evidence and gates needed for later review without exposing private credentials.

Evidence standards

Performance without performance theater

We separate system readiness, paper execution, and strategy evidence. A working agent is not automatically a proven edge—and the interface never pretends otherwise.

SystemCan it operate safely?

Runtime, data capability, broker connection, and refusal paths.

ExecutionWhat happened on paper?

Orders, fills, lifecycle events, and reconciliation outcomes.

ResearchHas the strategy earned promotion?

Versioned tests, costs, calibration, drawdown, and sample quality.

Ready to inspect the agent?

See every decision.
Trust every boundary.

Open the same Mission Control surface used to monitor ConvictionOS in paper mode.

Open Mission Control