Design, backtest, and deploy autonomous trading modules to the cloud. Strategies are built with LLMs and executed across stocks, options, futures, and crypto — through a single institutional control plane.
Every fill, reject, and risk decision is journaled, replayed, and graded — then fed to Davinci, ENTON's verifier-driven research model program. The model proposes. The evidence decides.
ENTON unifies research, simulation, and execution behind one API and one cloud deployment target.
Compose trading modules in TypeScript or natural language. The ENTON SDK exposes market data, order routing, risk, and portfolio state as first-class primitives.
Replay tick-level history across asset classes. Language models iterate on parameters, read performance reports, and propose the next experiment.
Promote a module to a managed runtime with a single call. ENTON handles orchestration, broker sessions, observability, and guardrails.
Describe the edge, the universe, and the risk envelope. ENTON generates a typed module, wires it to live market data, and validates it against historical ticks before a single order is routed.
// docs.enton.ai/examples/adaptive-es import { defineModule } from "@enton/sdk"; export default defineModule({ id: "example_adaptive_es_v1", universe: ["ES"], // CME E-mini S&P 500 futures broker: "tradovate", account: "paper:example", risk: { maxPositionUsd: 30_000, dailyLoss: 750, flattenAt: "15:45 ET", }, strategy: ({ bars, llm, risk }) => llm.plan` adaptive trend-follower on ES. retune length + atr multiplier on rolling volatility, size by conviction, trail the winner with atr(length). `, }); // $ enton deploy example_adaptive_es_v1 --env paper // illustrative sample · see docs for the full walkthrough
Market data is normalized on ingest; orders are routed broker-native on the way out. Nothing between you and the venue.
Live trading is the ground truth. Davinci turns every journaled decision into verifier-graded benchmark tasks — training a model to make stricter research calls, and to stop on weak ideas.
Every fill, reject, journal row, and account-state decision the platform executes becomes training data that exists nowhere else.
Davinci is graded by deterministic verifiers — replay parity, honest out-of-sample, attributed live P&L. Never by its own opinion.
Each session grows the benchmark. Each benchmark tightens the strategies. Tighter strategies produce cleaner evidence — and the cycle repeats.
Cheap checks first, expensive truth last. Nothing earns reward before attributed live P&L — and failed branches are published, not buried.
OAuth into your own brokerage and deploy directly. ENTON never custodies capital — your broker does.
US equities via Alpaca — smart routing, fractionals, extended hours.
route → alpaca · smart order router · fractional + extended hours
Multi-leg spreads with Greeks, IV surfaces, and expiry handling.
route → alpaca · multi-leg spreads · OPRA-priced greeks
CME micros and majors via Tradovate — session- and margin-aware.
route → tradovate · CME micros & majors · margin-aware sizing
24/7 spot via Alpaca, CME crypto futures via Tradovate — one runtime.
route → alpaca spot · tradovate crypto futures
The same runtime, data, and risk rails — reachable from the browser, the terminal, your pocket, and your codebase.
An AI trading agent — research markets, analyze your portfolio, and execute in plain English.
Open CodeAn institutional market terminal — live multi-asset data, charts, and depth.
Open TerminalDeployments, positions, and P&L in your pocket. Kill switch included.
App StoreA typed module SDK — market data, order routing, and risk as primitives.
Read the docsThe operating system for autonomous capital — research, execution, and verification in one loop. Built with working quant researchers and prop traders.