PRISM is your own quant desk — a fully automated trading system powered by 18 specialist AI agents across 120+ instruments and 22 brokers. Bull argues FOR. Bear argues AGAINST. A swarm jury votes. A 13-check risk gate decides. No spreadsheets. No staring at charts. No emotion-driven button-mashing. It runs on its own, 24/7 — within your rules, your risk limits, your broker, your strategies. Pause, override, or kill it anytime.
Instead of a team of human analysts, traders, and risk managers, PRISM runs AI agents that each specialize in one area. One watches the news. One reads the charts. One tracks whale wallets. One monitors macro data.
They feed findings into a structured pipeline that debates whether to trade, sizes the position, checks every risk limit, and — only if everything aligns — executes.
You configure the rules. You pick your broker. You set your risk appetite. PRISM does the work.
Every potential trade flows through 8 stages. Data agents pull. Analysts score. Strategies signal. Bull and Bear debate. A swarm jury votes. Risk gate vetoes. Or executes. Every step is logged.
Bull argues FOR. Bear argues AGAINST. This adversarial debate forces consideration of both sides — catching blind spots a single AI model would miss.
Four data agents fire simultaneously. Market Data pulls prices and candles. News scans Finnhub, NewsAPI, CryptoPanic. Macro monitors FRED and Finnhub for CPI, NFP, GDP. On-Chain tracks Etherscan, DeFi Llama, and Dune Analytics.
Six agents run in parallel. Fundamental reads earnings. Sentiment tracks Fear & Greed. Whale spots large wallet flows. Correlation measures cross-asset risk. Regime classifies the market. And then there's the one that does the heavy lifting:
Pure math. No LLM. What would take a human hours happens in seconds — for free.
Your active strategies — Python or PineScript — run against the analysis data. They produce BUY, SELL, or HOLD signals with suggested entry, stop-loss, and take-profit levels.
One LLM argues FOR the trade. Another argues AGAINST. A courtroom trial for every trade — forced consideration of both sides. The full debate transcript is saved for review.
Three different AI models vote independently. Two of three must agree. Using models from different providers ensures diverse perspectives — or run math-only mode and skip the LLM layer entirely.
Position size. Daily loss. Drawdown. Correlation. Margin. Max positions. Broker health. Slippage. Daily trade count. Cooldown. Event risk. Regime budget. Profit target. Any single failure vetoes the trade.
The order sends to your broker. Stop-loss and take-profit orders placed at the broker level — so they execute even if PRISM goes offline.
Fill verified. Price checked. SL/TP placement confirmed. Everything written to the database — every vote, every reason, every fill. Exportable, auditable, yours.
Each agent owns one narrow domain. None can force a trade alone. All votes are weighted by trust scores that update based on prediction accuracy — agents that are right more often gain more influence.
PRISM doesn't just run strategies. It discovers them. Two autonomous research systems run in the background, finding new ideas and iterating on your existing ones.
Strategy Discovery scans financial sources, extracts trading logic, auto-generates Python code, and backtests the result. Each discovery surfaces with a name, source, backtest metrics, and confidence score.
Deploy to live strategies with one click, or reject it. You set the research interval. Each cycle costs ~$0.50–$2 in LLM credits.
Start with any strategy. AutoResearch iterates on it — tests a variation, keeps it if it improves out-of-sample Sharpe, reverts if not, then tries another. Repeat across generations.
Run the Manual Lab (your idea, PRISM iterates) or the Autonomous Lab (set a budget, let it cook). Full experiment log with OOS Sharpe tracking.
Eight sub-tabs of advanced analysis. Everything you'd build yourself if you had six months — integrated, tested, ready.
Probability of profit. Median equity. Probability of ruin. Percentile distribution. Shows where your strategy lives on the outcome distribution curve.
COVID-19, Crypto Winter 2022, Flash Crash 2010, Volmageddon, Black Monday. See what your strategy would have done through every modern market trauma.
Splits data into windows. Optimizes on one, tests on the next unseen window. OOS much worse than IS = overfit. The definitive anti-curve-fit test.
Beta, momentum, value, size, volatility. See which factors your strategies lean on — and which are currently paying.
Measures how correlated your strategies are with each other. Low correlation = real diversification. Running 5 strategies that all buy the same dip = one strategy.
Per broker. Per session. Per instrument. Know exactly how much edge you lose at each venue — then route accordingly.
Live funding rates across crypto perps. Spot extreme funding — build funding arbitrage strategies on top.
How deep is the book? What's the average trade size? Know where you can actually get size filled without moving the market.
Every agent's vote. Every risk check. Every fill. Pause any agent, adjust any threshold, review any trade's full decision chain. Simple and Advanced modes — you pick the depth.
Trade forex, crypto, stocks, indices, metals, and oil — all from the same dashboard. Data auto-routes to the right provider for each asset class.
PRISM connects via your API keys with your permissions. It can trade — but it cannot withdraw. Revoke access anytime from your broker.
PRISM is a software tool, not a fund. It doesn't hold your money, doesn't manage it, doesn't make decisions for you. You configure. You monitor. You own the outcome.
PRISM always boots in simulation. Identical pipeline, identical analysis, identical risk checks — just no real orders. You must explicitly confirm to switch to LIVE each session.
One-click Kill Switch halts new trades and cancels pending orders. Dead-Man Switch auto-halts if the system stops heartbeating. Both visible in the header at all times.
YELLOW cuts position sizes 50%. ORANGE blocks new trades. RED halts everything. Thresholds are yours — default is 5 / 10 / 15% drawdown.
Bring your broker API keys — scope them trade-only (no withdrawal permission). PRISM never stores credentials anywhere other than your broker. You can revoke anytime.
Every agent vote, every risk check, every fill is written to your database. Filter, export, or replay any decision. If PRISM did something, you can see exactly why.
Skip the LLM layer entirely. Run pure rules-based strategies with zero per-cycle API spend. Keep PRISM's speed and structure without the AI bill.
PRISM always boots in simulation. Identical pipeline, identical analysis, identical risk checks — just no real orders. You must explicitly confirm to switch to LIVE each session.
One-click Kill Switch halts new trades and cancels pending orders. Dead-Man Switch auto-halts if the system stops heartbeating. Both visible in the header at all times.
YELLOW cuts position sizes 50%. ORANGE blocks new trades. RED halts everything. Thresholds are yours — default is 5 / 10 / 15% drawdown.
Bring your broker API keys — scope them trade-only (no withdrawal permission). PRISM never stores credentials anywhere other than your broker. You can revoke anytime.
Every agent vote, every risk check, every fill is written to your database. Filter, export, or replay any decision. If PRISM did something, you can see exactly why.
Skip the LLM layer entirely. Run pure rules-based strategies with zero per-cycle API spend. Keep PRISM's speed and structure without the AI bill.
Before any order leaves, the Risk Manager runs 13 independent checks. Any single failure vetoes the trade — no exceptions, no override.
14-day free trial on Pro. No credit card required. Cancel anytime. No hidden fees.
AI-Enhanced mode uses your own API key — OpenRouter or LiteLLM, your call. Typical usage runs about $5–$30/mo depending on how much you trade, billed directly by your provider (not us). Flip on Math-Only mode anytime to drop that cost to zero.
Private beta is limited. 14-day trial on either plan. Start on paper, go live when you're confident.
Request Beta Access →