Most strategies die here. That's the point.

Algo Operator is an agentic research framework. Six agents find, test and judge trading strategies, and one Critic decides which ones survive. Every verdict is logged, the rejections included.

Get access See how it works

The agent that pitches it never judges it.

  • findThe Scout turns papers, courses and quant blogs into testable candidates.
  • testThe Backtester runs each one on real tick data. The Analyst writes why it should work.
  • judgeThe Critic checks costs, concentration, parameter stability and held-out data. Passed, flagged or failed.
  • recordThe Registry logs every verdict. The Guardian makes sure held-out data is only used once.

Six agents. One verdict. Every result logged.

The agents do the research. You keep the pipeline, the logs and the final call on what goes live.

Get access Read the verdict log

Most of your trading week isn't trading.

Your edge is judgment: knowing your market, reading the regime, deciding what's worth testing. Too much of the week disappears into research busywork.

01 / 03
Finding ideasstage 1 · find
Your value
Knowing your market and spotting which ideas fit the way you trade.
Busywork
Reading papers and threads, turning vague rules into exact entries and exits, keeping track of what you already tried.

Taken over by Scout

Your value
Deciding which question a test should answer.
Busywork
Coding every variant, running backtests on tick data, copying profit factor, Sharpe and drawdown into spreadsheets.

Taken over by BacktesterAnalyst

Your value
Deciding what goes live, and how much you risk on it.
Busywork
Rechecking costs and parameter stability, keeping held-out data clean, logging every verdict so you never fool yourself twice.

Taken over by CriticRegistryGuardian

Most of your trading week isn't trading.

Your edge is judgment: knowing your market, reading the regime, deciding what's worth testing. Too much of the week disappears into research busywork.

Finding ideasstage 1 · find

Your value
Knowing your market and spotting which ideas fit the way you trade.
Busywork
Reading papers and threads, turning vague rules into exact entries and exits, keeping track of what you already tried.

Taken over by Scout

Testingstage 2 · test

Your value
Deciding which question a test should answer.
Busywork
Coding every variant, running backtests on tick data, copying profit factor, Sharpe and drawdown into spreadsheets.

Taken over by BacktesterAnalyst

Judgingstage 3 · judge

Your value
Deciding what goes live, and how much you risk on it.
Busywork
Rechecking costs and parameter stability, keeping held-out data clean, logging every verdict so you never fool yourself twice.

Taken over by CriticRegistryGuardian

Your edge shouldn't depend on your screen time.

Another course, another mentor, another weekend of backtesting, and the setup still breaks live. Your research grows, but it never gets faster or more reliable.

  • Get your weekends back
  • Built to catch curve-fits before they cost you
  • Every verdict logged, failures included

Build the system behind every trade.

Your knowledge, six agents and your broker live in one set of files. Switch between Claude, Codex or Cursor and every skill works the same, because the system is yours, not the AI's.

  • Claude
  • Codex
  • Cursor

Agents

  • Scoutfinds ideas
  • Backtestertests on tick data
  • Analystargues the case
  • Criticthe only verdict
  • Optimizerrefines what runs
  • Guardianguards held-out data

Your Trading OS

Same skills. Same workflows.
Any provider.

Connected

  • Darwinexbroker
  • MetaTrader 5execution
  • Dukascopy ticks9+ years of data
  • Registryverdict log
  • Telegramalerts

Built on a simple framework:
Context, Tools, Skills.

Context

What your agents need to know

Your markets, sessions, risk rules, trading style and everything you already tested.

Tools

Where your agents can work

Tick data, the backtester, MetaTrader 5, your broker, the verdict log and alerts.

Skills

How the work should be done

Your repeatable research steps, checks and standards, written down once and run every time.

Build my Trading OS

Courses teach validation. Builders sell the tools. This has a judge.

Typical courseTypical strategy builderAlgo Operator
Who decides a strategy is goodYou, after the courseYou, reading the reportsAn independent Critic, on fixed checks
What happens to ideas that failRarely shownDiscardedLogged with the reason, permanently
Held-out dataExplained onceAvailable as a settingGuarded, and only spent once
What you are soldThe methodThe softwareThe research line and its record

The verdict log is the proof.

Every verdict lands in an append-only log with its date, source and reason. The rejections stay in. A high failure rate means the gate is doing its job.

Candidates tested
—
Failed
—
Passed
—

Live counts appear here when the public log opens.

Example entries. These show the format only and are not real results.

FAILED

71% of the edge came from 3 trades.

concentration check · source: public quant blog

FLAGGED

Edge disappears when the spread is 1.5× wider.

cost check · source: Scout

FAILED

Neighbouring parameter values collapse the result.

parameter stability · source: course material

PASSED

Held the held-out window after costs.

all checks · moves to the next stage, not to your account

Bring a strategy. Let the Critic look at it.

Request access and tell us what you trade. Access opens in small groups, so every new operator gets a proper start.

  • Research software you run. No signals, no managed accounts.
  • You decide what goes live, always.
  • No affiliate links to brokers or prop firms.

Questions

Is this a signal service?

No. Algo Operator is research software. It finds, tests and judges strategies. It does not tell you what to buy or sell, and it does not trade on your account.

Do you promise returns?

No. Backtests are simulated results and are not a prediction of future performance. We show verdicts, reasons and methodology instead of profit claims.

What does the Critic check?

Trading costs, how concentrated the result is, how stable it is when parameters move, and how it holds up on data it has never seen. The outcome is passed, flagged or failed.

Who decides what goes live?

You. A passed verdict means a strategy may move to the next stage. It is never an instruction to trade.