Why ApexAlgorithmic trading you can defend.When a trade costs real money, "the strategy said so" is not a defense. Apex is the trading engine that pairs battle-tested sub-strategies with full audit: every signal recorded, every gate verdict replayable, every broker call linked back to the decision that produced it. And the audit chain runs both ways. Your own notes on past trades feed back into future signals, so the engine learns your reasoning, not just its own.

The piecesFive Rust crates, one HTTP surface, one CLI. Each crate owns one concern; the engine composes them into a complete trading loop.

Strategies emit; gates decide; broker executesapex-coreThe engine that orchestrates the loop. Strategies are values, not types: ten built-in sub-strategy rules plus a trait you can implement. Gates compose in registration order; the first rejection stops the chain. Every step lands in the Cortex provenance graph.

Ten battle-tested sub-strategy rulesapex-strategiesVWAP reclaim, opening-range breakout, bull-flag continuation, failed-breakdown reversal, higher-lows continuation, squeeze-to-expansion, trend-pullback buy, volume breakout, volume dry-up reversal, volume exhaustion reversal. Each rule has clear preconditions and a documented predictive premise.

Plug in your broker; the contract stays the sameapex-brokerTrait-based broker abstraction. Alpaca is in; IBKR, Schwab, Robinhood, and crypto venues land behind the same Broker trait. WebSocket trade-updates auto-reconnect via helix-rust's reconnecting_stream so a transient network blip never corrupts position state.

Replay any window; pin behavior with a canaryapex-backtestHistorical replay with configurable fill model (current-bar-close or next-bar-open), slippage in basis points, multi-symbol bar streams, walk-forward parameter sweeps, IC scoring, and a JSON behavioral canary that pins per-strategy emission so refactors flip red on drift.

B2B-API-first; one router, one daemonapex-apiAxum HTTP surface exposing the engine. Discovery (`/strategies`), backtest (`POST /backtest`), live monitoring (`/live/*`), and order submission (`POST /orders`) with idempotency keys. Optional Bearer auth gates everything except `/health`. Run as a library; embed in apex-cli serve; future-proof for a standalone daemon.

How it composesA bar arrives. Strategies emit signals. Gates approve or reject. The broker submits the survivors. Every step lands in the audit chain.

Battle-tested rulesEach sub-strategy is a single, named rule with a documented predictive premise. No weighted-score aggregates; no opaque ensembles. vwap_reclaim reads sentiment when wired; the others stay pattern-based. New rules slot in behind one trait.

Risk gates in seriesCircuitBreaker halts on consecutive losses, DailyLossGate caps drawdown, OpenPositionsGate bounds exposure. Gates compose in registration order; first rejection stops the chain. Every verdict records a provenance node with the gate id and the reason string.

Audit chain by constructionEvery signal, gate verdict, and broker call writes a Cortex provenance node with parent edges back through whatever fed it.cortex query <node-id> walks from a filled order back through the sentiment that informed it. The audit chain is the defensibility promise.

Your reasoning, fed back inNotes you write on trades become first-class nodes in the same graph. Cortex embeds, indexes, and surfaces them at decision time, so the next vwap_reclaim on a thinly-traded small cap comes with your six prior notes on exactly that setup. An opt-in TraderConfidenceGate turns the aggregate sentiment into a rejection signal you control. The system records what it did; the journal records what you learned; both inform the next trade.

Where it fitsB2B-API-first GTM. Integrators build trading products on apex; end-user SaaS is a follow-on. The audit chain is what makes the engine defensible across every deployment model.

Why did this trade fire?Solo algo traderYour keys, your broker, your trades. The engine runs on hardware you control; the hosted dashboard at cloud.apex.fusion.dev is the operator surface. Every entry traces back through the signal, the gate, the indicators, and the sentiment that triggered it. When a trade hits a stop, the decision chain is one query away.

Can our model survive walk-forward?Quantitative researchRun a parameter grid through apex-backtest with per-symbol assignments. IC scoring tells you whether confidence correlates with the trades that mattered; the behavioral canary pins per-strategy emission so a refactor that silently shifts signals flips red.

How do we let customers trade through us?Fintech integratorStand up apex-api against your tenant model. Per-tenant audit trails through Cortex, broker-agnostic execution behind the Broker trait, optional Bearer auth. Phase 4 adds per-tenant rate limits and scoped tokens for cloud deployments.

Where's the evidence?Compliance + auditWhen a regulator, auditor, or risk officer asks why a position was opened, the answer is a node id. Walk back through the chain: broker submission, gate verdict, signal decision, the bars and sentiment that fed it. Append-only; tamper-evident.

Ready to build on it?Open source for individual traders. Alpha for fintech integrators who want a trading engine they can audit, replay, and defend.

Read the docsRequest alpha access