Autonomous strategy
development.

Describe a trading idea in plain English. The lab builds it, tests it on real tick data, and tells you the truth about whether it works.

NQ · 2026-06-01 · 5m replayLive replay

Real NQ session · 2026-06-01 · 5-minute bars built by the AlgoLab engine from its own tick data — not a mock-up.

What's under the hood

A lab, not a chart toy.

0+ real NQ sessions

Quote- and trade-level market data, replayed tick by tick. Your idea trades the actual tape — every fill, every fee, every unlucky moment included.

Plug it into Claude Code over MCP — your agent, our engine. The same honest backtests, driven by whatever is doing the thinking.

Deflated Sharpe, trial counting, and minimum-sample gates. Every retry you make is counted against the result — the math can't flatter you.

BUY/SELL markers and TradingView alerts, generated from the exact card that was tested. The model never freestyles your script.

in development

Built for the eval era: a daily risk governor and eval-aware execution are in development — the lab already grades ideas with that goal in mind.

No code. No Python. No spreadsheets of copied numbers. You bring the idea in plain English — the lab runs the math, and it can't lie to you.

How it works

01
Describe it
Say your idea like you'd tell a friend — the lab turns it into an exact strategy card you approve before anything runs.
02
Test it on real tape
Your card replays through hundreds of real recorded sessions, tick by tick, and gets graded: edge or luck.
03
Follow it on TradingView
Export the same rules as a Pine script — buy and sell markers plus alerts, nothing added, nothing improvised.