Shape your idea before coding
Use Strategy Builder to clarify the market and rules, inspect feasibility and prior research, and approve or revise the hypothesis before development.
Explore Strategy BuilderYour AI trading strategy factory.
Turn ideas into trading strategies with a coordinated team of AI agents. Stratmill automates coding, backtesting, parameter optimization and risk review, with paper trading and separately authorized live execution.
Strategy development, validation and controlled execution.
strategy.pyclass Alpha(FactoryStrategy):
def calculate_signal(self, bar):
fast = self.fast_ma.value
slow = self.slow_ma.value
return (fast - slow) / slow
Illustrative example · not a live result
AI agents build and refine your strategies.
An AI research agent develops the idea. An AI coding agent builds and revises the strategy.
Explore strategy researchEvidence before conviction.
An AI analyst examines backtests. A separate AI risk reviewer challenges the findings.
Explore backtesting
Forex price and economic data sources
Yahoo Finance
Dukascopy
BIS
ECB
Inside the AI strategy factory
One factory. From idea to execution.
01
Generate ideas
Start with your own idea or enable Autopilot. The AI research agent develops testable hypotheses using available market context, prior outcomes and recorded lessons.
02
Build & verify
The AI developer writes and revises the strategy. Automated verification and an AI QA reviewer check the code, data use and trading logic.
03
Backtest
The backtest engine simulates trades with historical data, configured fees and execution assumptions. The AI analyst evaluates the results and recommends further testing, revision or abandonment.
04
Optimize & review
Optimization tools run sensitivity, walk-forward and holdout tests. The AI analyst assesses robustness; a separate AI risk reviewer examines the evidence and proposed exceptions.
05
Paper & live
The AI portfolio agent reviews promotion and portfolio fit. Approved strategies enter paper trading; supported live execution requires your authorization, a connected account and execution controls.
Follow the AI agents’ work through recorded stages, reports and review decisions. A failed check or review can send the strategy back for revision or end the run.
See the strategy factory in motion.
Illustrative example · not a live result
Watch AI agents build and review a strategy.
AI agents develop the idea, write code and assess the test results. Review findings can send the strategy back for another iteration.
Watch the path, not just the endpoint.
An illustrative equity curve reveals setbacks along the way. Inspect drawdown and trading costs before interpreting a result.
Keep open positions in the picture.
Compare account equity with closed-trade balance. An open position can change equity before a trade closes.
- Describe
- Build & verify
- Backtest
- Stress-test
- Paper-trade
EquityClosed-trade balanceDrawdown
Net return—
Max drawdown—
Checkpoint—
More ways to develop your strategy
Bring your code and data
Import code you have the right to use, or validate your own dataset in My Data. Compatibility, source coverage and translation differences stay visible.
Explore imports and dataTest, compare and export
Run a separate experiment in the Backtest Lab, inspect its parameters and results, and review target-specific compatibility before exporting a strategy.
Explore the Backtest LabInside the product
Open a real research record.
This public strategy was rejected after review. Inspect its hypothesis and recorded findings to see why a positive headline return is not enough.
Read the public report
Bring your strategy. Take your code.
Import, review and export across platforms.
Start with code you own, or export a strategy you can access. Each translation shows its compatibility limits and the checks actually performed.
Import from ten source formats
TradingView Pine Script, Freqtrade, Jesse, backtrader, QuantConnect LEAN, MetaTrader 4, MetaTrader 5, FMZ, vn.py and Hummingbot. Upload up to 20 files and review the credit estimate before starting.
Explore strategy importsReview the translation
Your original code stays private. Review changes in indicators, orders and available data, then inspect verification and any trade comparison. Importing code never places live orders.
Read the import guideExport to eight platforms
Download compatible ports for backtrader, vn.py, TradingView, QuantConnect LEAN or MetaTrader 5. Each report identifies the checks actually performed; a structural self-check does not establish trade parity.
Compare export checksOne factory, focused on Forex
Research the 32 configured currency pairs with source-aware history, FX sessions and recorded costs. Research, paper observation and broker execution have separate readiness checks.
Currency pairs
Develop ideas for majors and crosses. Every traded and signal-only instrument belongs to the same Forex universe.
Explore the Forex universeSession-aware research
Use closed bars, synchronized currency pairs and daylight-saving-aware session calendars. Missing history stays visible.
Recorded trading costs
Inspect commission, spread and overnight financing assumptions. A research cost model is not a broker settlement record.
Controlled execution
ZFX profiles and MT4/MT5 bridges support broker onboarding. Demo connectivity and captured account terms must pass before execution can be admitted.
What you actually get
A full backtest report, not a signal
The backtest engine produces performance, drawdown and trade records. AI agents review that evidence and record their findings. This example illustrates the report format; open the strategy list to inspect published records.
Illustrative example · not a live result
EURUSD trend research · illustrative reportHoldout passed
Strategy
Buy & hold
Illustrative data · modelled spread, commission and financing
Risk & performance
+147%
Total return
2.31
Sharpe
1.86
Sortino
-12.4%
Max drawdown
68%
Win rate
2.68
Profit factor
Per-trade return distribution — positive skew after fees
Don't take our word for it — these pages are open, no account needed:
No fantasy fills
Costs belong in the backtest
FX costs include commission, spread and overnight financing. Inspect the frozen assumptions recorded with each backtest and test sensitivity to wider spreads.
Scroll horizontally to see all columns.
| Forex cost component | Research treatment |
|---|---|
| Spread and commission | Configured pair costs frozen with each evaluation. |
| Overnight financing | Approved policy-rate evidence and modelled rollover charges. |
| Currency conversion | Quote-currency accounting with available conversion evidence. |
| Broker terms | Captured separately from the designated demo terminal. |
These are simulation inputs, not broker quotes. Instrument-specific fees, financing, spread, slippage and impact can change the result. Inspect the run’s cost model and data coverage before comparing strategies.
Built for rigor
AI agents working through one connected pipeline.
A team of AI agents
AI research, coding, QA, analysis, risk and portfolio agents have distinct roles. The factory connects their decisions to testing engines, revision loops and release checks, keeping the strategy’s development history together.
Overfitting defense
Automated verification, walk-forward and holdout tests probe overfitting. The AI analyst evaluates the findings, and a separate AI risk reviewer examines proposed exceptions before promotion.
Real-market microstructure
Historical market data, configured fees and execution assumptions model trading costs. Coverage and fill realism depend on the instrument and available data.
Strategy marketplace
Inspect published strategies and their backtest records before subscribing. Publishers receive 70% of strategy subscription revenue.
Inherited research corpus
AI research agents can draw on prior outcomes and recorded lessons. Paid workspaces can also use the platform research corpus; private workspace research stays separate.
Backtest–paper parity
Paper trading runs the exact same engine as the backtest — each survivor in its own isolated live node, filled against real exchange quotes.
Venue-managed exits
Strategies can attach stop-loss and take-profit as venue-side bracket orders — the same contingent orders a desk would place, modeled end to end.
Mark-to-market truth
Equity curves are marked to market on every bar, unrealized PnL included — and leveraged strategies get liquidated when margin runs out, just like on the venue.
Deep market data
Explore source-owned FX history and separately validated macro observations. Check the Data page for coverage and availability before testing an idea.
Marketplace
Explore strategies and inspect the evidence
Explore Forex research. Review each published strategy’s trades, costs, drawdown and validation before subscribing. These examples are ideas, not listings or results.
Browse the marketplaceIllustrative example · not a live result
Session momentum research
Compare funding accrual, fees and hedging costs.
Currency relative-value research
Compare synchronized currency pairs using available history and FX session calendars.
Forex trend research
Inspect spreads, overnight carry and validation across currency pairs.
Forex breakout research
Inspect XAU/USD trades, session rules and financing.
Explore more of the strategy factory.
Follow the workflow from forecast research to portfolio decisions and controlled execution. Each stage has its own evidence and responsibilities.
Strategy research
Idea development is the first stage of the factory. AI research and coding agents turn a hypothesis into strategy code; tests and reviews guide later revisions.
Explore strategy researchKnowledge library
The curated research corpus that Stratmill's AI agents read: books, papers, notebooks and open-source trading code, with their figures. Search it, read it, and send any document into the research pipeline.
Explore the knowledge libraryAlpha & signals
A signal expresses direction and conviction. Position sizing, rebalancing and execution determine how that forecast becomes a trading strategy.
Explore alpha & signalsPortfolio construction
A correlation matrix reveals strategies that move together. Effective bets help distinguish the number of strategies from the amount of independent exposure.
Explore portfolio constructionAutomations
Automations connect Stratmill to tools outside it, starting with TradingView alerts and strategy scripts. Each automation asks for your informed consent first, and you can revoke it at any time.
Explore automationsLive trading
Live authorization is tied to a strategy release and your connected account. A paper result or a plan purchase alone does not authorize real orders.
Explore live tradingPricing
Start free. Scale when it works.
Credits meter the AI work your research consumes. Start with a free monthly allowance; paid plans add capacity and access to the platform research corpus.
Free
$0
- 500 credits / month
- 2 active pipeline runs
- Unlimited paper trading
- Learns from your own runs
Most popular
Pro
$99 / month
- 8,000 credits / month
- 8 active pipeline runs
- Unlimited paper trading
- Inherits the platform research corpus
Max
$399 / month
- 40,000 credits / month
- 20 active pipeline runs
- Unlimited paper trading
- Inherits the platform research corpus
Active runs cap how many strategies are in research at the same time; they are not a daily launch allowance. Daily launch limits and available credits also apply.
Questions
What you're actually signing up for
Can the AI strategy factory generate its own ideas?
Yes. When you enable Autopilot, the AI research agent can propose new hypotheses and start them through the development pipeline. It uses the context and prior outcomes available to your workspace. Credits, daily launch limits and available capacity still apply. You can also submit your own ideas, and every strategy remains subject to testing and review.
What are Stratmill’s AI agents?
They are AI-powered software agents, each assigned a role in strategy development: research, coding, QA, analysis, risk review or portfolio review. Backtesting and verification engines run the simulations and automated checks that inform those reviews. Strategies must pass the applicable admission and release checks before paper trading; an AI recommendation alone does not enable live trading.
Do I need to know how to code or trade?
You can start in plain language. An AI research agent develops your hypothesis, an AI developer writes the code, and AI QA and analyst agents review the implementation and results. Trading knowledge helps you judge their work, assumptions and risks.
Is real money ever at risk?
Research and paper trading use simulated funds. Live execution is a separate, restricted workflow requiring explicit authorization and a configured trading account. It is not enabled by starting a research run or buying a plan, and real capital can be lost if live trading is enabled.
How do you prevent overfitting?
We use sensitivity analysis, walk-forward and holdout tests, plus checks such as deflated Sharpe and backtest-overfitting probability. Validity gates and risk review govern promotion; some quality thresholds allow reviewed exceptions. These checks cannot guarantee future performance.
Are the backtests realistic?
Backtests use configured venue fees and models for fills and applicable financing and margin costs. Review the data coverage and assumptions with the report: historical simulations cannot reproduce every live execution condition.
Which markets can strategies trade?
This factory supports only the configured Forex currency pairs. Check historical coverage, the recorded cost model and independent paper or broker readiness before choosing a strategy.
What are credits?
A prepaid unit of platform usage that meters the AI compute your research consumes. Free plans include a monthly allowance; paid plans include more and never-expiring packs are available.
Can I start without a payment card?
Yes. You can create a free account without a payment card. The free plan includes a monthly credit allowance, concurrent research capacity and unlimited paper trading. The pricing page lists the current allowances and the options for upgrading.
What can I explore before creating an account?
You can browse public market pages, published strategies and marketplace listings. Read the available reports and evidence before starting your own research. Workspace actions and access to strategy code depend on sign-in, ownership and the applicable access rules.
Will every research run produce a profitable strategy?
No. A run may need several revisions or be abandoned after checks or review. Credits meter the AI work performed, so a run can consume credits even if it never reaches paper trading. The useful output may be evidence that an idea should not be pursued.
Do unused monthly credits carry over?
The monthly allowance resets each cycle rather than accumulating. Purchased credit packs do not expire and are spent after the monthly allowance. Credits are separate from concurrent-run limits; adding a pack does not change those plan limits.
Can I inspect the strategy’s source code?
Source-code access depends on ownership and the strategy’s sharing or marketplace rules. An available public report does not necessarily include unrestricted code access. Review the access offered for the particular strategy before subscribing to or using it.
How long does a research run take?
There is no fixed completion time. Data requirements, simulation work, agent availability, revisions and reviews all affect the duration. Follow the run’s stage and history in your workspace; starting a run does not guarantee a strategy will be approved.
The math behind every verdict
Make weaknesses visible before promotion
The AI analyst and AI risk reviewer examine these checks alongside the strategy report. Validity failures block promotion; quality findings feed into their assessment, including any documented exceptions.
Deflated Sharpe ratio
Checks for performance inflated by repeated trials. Deflated Sharpe accounts for trial count and return characteristics.
Walk-forward optimization
Checks for parameters tuned to the past. Fit in-sample, scored out-of-sample across rolling windows — decay shows up immediately.
One-shot holdout
Checks for overfitting to the whole dataset. A final untouched slice is tested exactly once — no second chances, no peeking.
Sensitivity & cliff detection
Checks for fragile parameter choices. Sensitivity sweeps show how nearby settings change performance for review.
Cost-clearing gate
Checks for returns too small to cover modeled trading costs. Cost checks are evaluated with the other promotion evidence.
IS/OOS decay ratio
Checks for in-sample flattery. If out-of-sample performance falls too far below in-sample, the result is treated as noise.
Know what the evidence can tell you.
Check the inputs
Data availability, instrument history, fees and fill assumptions vary by venue. Review the coverage and assumptions recorded for your run.
Explore market dataSeparate tests from forward evidence
Backtests simulate the past. Paper trading adds forward observations. Low-confidence results are not statistically proven, and neither is a promise of returns.
Explore paper tradingAuthorize live trading separately
Opening a workspace, importing code or starting a backtest does not authorize real-money orders. Live execution requires your approval and a supported connected account.
Explore live tradingStart your AI strategy factory.
Create your first strategy run on the free plan. Follow the AI agents through code, tests, revisions and review, then monitor strategies admitted to paper trading.
Free plan · no card required · simulated paper trading
Stratmill is an AI trading strategy platform, not financial advice or a broker. Backtest and paper results are hypothetical and not indicative of future performance. Figures shown are illustrative. Trading involves substantial risk of loss.