Documentation
VeriRun Lab documentation
Everything you need to take a futures-strategy idea from a plain-English sentence to an honestly graded backtest — and to keep your research organized along the way. Every guide is written end-to-end: follow one top to bottom and you will have done the real thing in the app.
The research loop at a glance
VeriRun Lab is built around one loop, and the docs follow it in order:
- Describe or write a strategy — with the AI Strategy Builder or by hand in the workspace IDE.
- Test it. A required suite of tests proves the code does what the strategy says. Nothing runs on data until the tests are green.
- Preview it. A short realistic run over a few sessions shows you what the strategy actually does on a chart — the same fills a backtest would give.
- Backtest it. A conservative, research-grade run produces an A–F scorecard with a pre-committed GO / NO-GO verdict.
- Keep the evidence. Every run lands in your library — pin it, annotate it, compare it, export it.
Getting started
Create your account, sign in, make your first project and workspace. Ten minutes, start to finish.
Sign-in & security
Passkeys, authenticator apps and recovery codes: the second step every sign-in asks for, and what we store.
AI Strategy Builder
Describe an idea in plain English, answer clarifying questions, review the Brief and the generated code — and stay in control of every change.
AI Assistant
One conversational surface on every page: ask questions, attach images, or have it change code — it reads, proposes a diff, runs the checks, and applies nothing until you accept.
Writing strategies by hand
The workspace IDE: what every file does, the strategy contract, the market data available to your code, and snapshots you can always roll back to.
Indicators & features
The full reference: around 48 technical indicators, pivot systems, and a deep order-flow suite (footprint, delta, value area, absorption) your strategies can read.
Testing your strategy
The twelve test categories, how to run your tests, and when a failing test holds Preview and Backtest back.
Preview & Backtest
A short realistic preview versus the full graded backtest, how to launch either, and how to read the chart and the results.
The scorecard
What A–F measures, why a NO-GO caps the grade, and why an honest NO-GO means the system is working.
The run library
Pin, annotate, compare and export runs — and read the regime, stress and price-path robustness views correctly.
Optimization
Search your strategy's parameters with a study — scored on the train window only, the test window shown but never scored, and a multiple-testing verdict on every rank.
Screener
Fan one approved strategy across many instruments and timeframes in a single run, then rank the results — with correlated markets counted once.
Notebooks
Free-form research code next to your strategy, including a research backtest call — and the limits it runs under.
Meta-labeling
Train a secondary model to gate your signals, then read it out-of-sample: precision, recall, lift, ROC-AUC, feature importance, and a precision-vs-coverage sweep.
Journal & firm tools
Log trades, import fills, attach screenshots, build playbooks and goals, track prop-firm rules and evaluations, and replay real trades through a firm's payout math.
Sharing & the leaderboard
Publish a run once its audit evidence is in place, and read a board ranked by that evidence — never by raw return. Flagged runs publish too, flag showing.
External agents (MCP)
Drive the platform from your own AI coding agent over MCP: a read-first research surface with scoped machine tokens and human-accepted draft edits.
Team plugin (Claude Code)
One command installs four role-scoped agents — researcher, runner, audit-reader, chart-reader — over the gated MCP façade. Approval stays human and in-app.
Data & credits
What the data is, how usage and credits are shown, and how to read the cost estimate before you launch a run.
One thing to know before you start
VeriRun Lab is deliberately conservative. Backtests fill your orders under pessimistic rules, the grading gates are fixed before a run, and a strategy that fails them gets a NO-GO — clearly, and on purpose. If you are used to backtesting tools that flatter you, some verdicts here will feel harsh. That is the product working as designed: a result you can trust is worth more than a result you like.
Research tooling, not trading advice
Everything the platform produces is a research result — probabilities, not promises. Nothing in the app or these docs is a recommendation to buy or sell any instrument.