Strategies that prove
they are profitable

Constellation is an algorithmic trading engine that uses on-board AI to evaluate, validate, and run trading strategies. No strategy touches your capital until it has proven itself through rigorous backtesting, a 20-day paper proving period, and continuous AI review.

The problem

Most trading algorithms fail in production because they were never truly tested.

Backtesting on perfect data with no slippage, no spread, and no market impact gives you a strategy that looks profitable on paper but bleeds money in reality. Curve-fitted parameters. Overfitted signals. Results that vanish the moment real capital is at stake.

Constellation was built to solve this. Every strategy must survive realistic conditions, prove itself with live market data, and pass AI review before it earns the right to trade.

How it works

Four stages between your strategy and your capital.

01

Build or install a strategy

Create your own strategy from parameterised templates or install one from the community library. Define your risk limits, position sizing, and the markets you want to trade.

02

Realistic backtesting

Your strategy runs against historical data with ADV-bucketed slippage, gap-through stops, and spread modelling. Anti-gaming penalties catch overfitted parameters. Only strategies that demonstrate genuine edge survive.

03

20-day paper proving

Strategies that pass backtesting enter a paper trading period using real-time market data. Performance is tracked against the same metrics as live trading. If results degrade, the strategy is automatically demoted.

04

Live trading with AI guardrails

Every trade passes through five authorization gates before execution: strategy conviction, portfolio concentration, drawdown limits, market conditions, and an on-board LLM review that validates the trade thesis and provides a confidence score.

On-board intelligence

An LLM that watches every trade.

Constellation runs a local large language model as part of the execution pipeline. Before any trade is placed, the LLM reviews the strategy's thesis, validates current market conditions, checks for concentration risk, and returns a confidence score. This isn't a chatbot. It's an autonomous review layer that operates within the trade decision loop.

Trade review

Every entry and exit is validated against the strategy thesis and current market regime.

Strategy proposals

The LLM analyses performance data and proposes parameter adjustments with evidence and expected improvement.

Risk monitoring

Continuous assessment of portfolio concentration, correlation, and drawdown trajectory across all lanes.

Platform

Everything runs on your machine.

Native macOS app

The Observatory is a native SwiftUI app with a star-map interface for monitoring your strategies, satellites, and positions in real time. Your server, your data, your keys.

Multi-lane isolation

Run swing trading, day trading, and crypto strategies simultaneously across isolated lanes. Each lane has its own elected champion, risk budget, and independent kill switches.

Strategy marketplace

Browse and install community strategies from the library. Every published strategy includes full backtesting results, parameter documentation, and version history.

Automated pipeline

Once configured, the pipeline runs autonomously: daily backtests, champion re-election, trust progression, and execution. Manual intervention is always available but never required.

Stop guessing. Start proving.

Create an account and deploy your first strategy. Free tier includes full backtesting access and community library browsing.