No-code algorithmic trading

Turn a trading idea into a tested strategy.

Describe your idea in words. Numir's AI turns it into a visual strategy you can adjust node by node, then backtests it against years of data. No Python, no libraries to wrangle.

Runs on the same engine that would trade it live
EUR/USD trend follower · valid

let sma_fast be the SMA of close, period 96

let sma_slow be the SMA of close, period 480

when sma_fast is above sma_slow and the EUR/USD position is flat

buy 200000 units of EUR/USD at market

when sma_fast is below sma_slow and the EUR/USD position is long

close all of the EUR/USD position

✓ backtest complete 49,778 bars net +$16,046.37

Algorithmic trading has always been locked behind code.

You have the market sense. What you don't have is a spare year to learn Python, pandas, and a backtesting framework just to test a single hunch.

So the idea stays in your head. Or in a spreadsheet that can't do anything with it. Numir is where you finally get to test it.

From a sentence to a backtest.

STEP 01

Describe

Type the strategy the way you'd explain it to a friend. "Buy euro-dollar when the day-long average crosses above the five-day one, sell when it crosses back."

STEP 02

Compose

The AI builds a node graph: signals, conditions, sizing, exits. Every block is visible, and you can change any of it.

STEP 03

Backtest

Run it against years of historical data and read the equity curve, drawdown, and win rate. Then adjust it and run it again.

The builder

A canvas that shows its work.

Every part of your strategy is a labelled block you can read and change. No hidden logic. Nothing you have to write in code.

  • Readable nodes.
    Indicators, conditions, position sizing and exits, each a labelled block instead of a formula you have to memorise.
  • Change anything.
    Edit a period, flip a condition, add a branch. The result updates as you work.
  • Test variations.
    Fork any strategy, try a variant, and keep the version that performs better.
EUR/USD trend follower · a real strategy, redrawn to scale 31 nodes · valid
every M15 · EUR/USD
◆ signal

sma_fast = SMA of close, period 96

◆ signal

sma_slow = SMA of close, period 480

◇ decision

when sma_fast is above sma_slow and the EUR/USD position is flat

● order

buy 200000 units of EUR/USD at market

◇ decision

when sma_fast is below sma_slow and the EUR/USD position is long

● order

close all of the EUR/USD position

✓ compiles · 31 nodes · edits auto-apply · ⌘Z undoes

Numir · assistant · a real session, redrawn to scale

swapped the rsi mean reversion for a trend-following sma crossover. added 96 and 480-period smas to capture the 24-hour and 120-hour trends on the 15m chart with a 200k unit size.

✦ 7 edits applied
✓ backtest complete 49,778 bars 258 fills net +$16,046.37 sim · dv-2y-m15
AI assistant

Describe it in words, get editable rules.

The assistant edits the strategy itself, not a summary of it. Its changes land on the canvas as plain sentences you can read and undo. The exchange shown here really happened, and the backtest behind it ran over two years of EUR/USD data.

It won't trade anything on its own. Every proposal has to compile against the engine's grammar before it touches your strategy.

See how it would have performed.

Every strategy runs against real historical data, on the same engine that would trade it live. You get the full equity curve, drawdowns included. It shows what would have happened to your account, trade by trade.

EQUITY — SMA cross · EUR/USD · Jul 2024 – Jun 2026 · 1h +0.0%
Total return
+16.9%
Win rate
38.5%
Max drawdown
−8.1%
Sharpe
1.64

An actual run from the Numir engine: an SMA-cross strategy over 12,450 hourly EUR/USD bars, 65 closed trades. This is an in-sample backtest. Past performance does not guarantee future results, and every strategy carries risk of loss.

Coming Q3 2026

Start with a sentence.

Stop translating your ideas into code someone else wrote the rules for. Describe a strategy. Find out whether it holds up.

Numir opens in private beta in Q3 2026, by invitation only. The first invites go to people we know, and every member gets a few codes of their own to pass on.