Build and test strategies Build and test strategies on institutional market data.

Institutional tick data with bid and ask, from minute bars down to every single tick with its size at the best price, plus server compute and an assistant that builds and backtests against that tape, not against guesswork.Server compute and an assistant that backtests against the tape, not guesswork.

From zero code to a backtest at full speed.

Let the agent handle the trial and error. It writes the strategy, validates it and fixes what fails, while you look for the next idea.

Orion Workspace

README.md

filters.py

strategy.pynew

notebook.md

09:41:03Read the workspace. Reused in_session from filters.py.

09:41:05Wrote strategy.py, LondonBreakout, 28 lines.

09:41:06Validated. 1 class, no import refused.

09:41:08Run started on EURUSD, tick tier.

09:41:10Report ready.

filters.py

strategy.py

ValidateRun

1from filters import in_session

4class LondonBreakout(Strategy):

5 """Buy a break of the Asian range in London."""

7 def __init__(self, params):

8 self.atr_period = int(params.get("atrPeriod", 14))

9 self.reward = float(params.get("reward", 2.0))

10 self.day = self.high = None

11 self.done = False

13 def on_bar(self, ctx):

14 bar = ctx.bars(1)[-1]

15 if bar.ts.date() != self.day:

16 self.day = bar.ts.date()

17 self.high, self.done = bar.high, False

18 if in_session(bar.ts, "asia"):

19 self.high = max(self.high, bar.high)

20 return

21 if self.done or not in_session(bar.ts, "london"):

22 return

23 atr = ctx.indicators.atr(self.atr_period)

24 if atr and bar.close > self.high:

25 stop = atr / ctx.pip()

26 target = stop * self.reward

27 ctx.signal("LONG", sl_pips=stop, tp_pips=target)

28 self.done = True

Python · 3 filesEURUSD · Tick · Jan to Jun 2024Report ready

London breakout

Build a London breakout for EURUSD. Take the high of the Asian range, buy the first close above it after 07:00 UTC, one ATR stop, 2R target. Use the session helper in filters.py.

Read workspace3 files

filters.py already has in_session, so I will import it rather than repeat it, and write the strategy against the engine's API.

Writing strategy.py+28

+ target = stop * self.reward

+ ctx.signal("LONG", sl_pips=stop, tp_pips=target)

+ self.done = True

Validate

1 strategy class: LondonBreakout
Imports: filters, from this workspace. None refused
Parameters: atrPeriod, reward

Valid, nothing refused. Run it on EURUSD at the tick tier and every fill uses the bid and ask as they were.

Run backtestSaveOptimize

Ask a follow up

Claude Opus 5

Describe it, get a strategy

Explain the rules in your own words in the assistant beside the editor. It answers with a strategy class for this engine, and one click puts it in the file.

Get early access

Five languages, one engine

Write it in Python, C# or Pine Script, or put it together from blocks, and read any of it as the Rust the core runs. Validate checks each one before the first bar is read.

Get early access

strategy.py

Python

if bar.close > self.high:

stop = atr / ctx.pip()

ctx.signal("LONG", sl_pips=stop)

Breakout.cs

if (Bars.ClosePrices.Last(1) > high)

ExecuteMarketOrder(TradeType.Buy,

SymbolName, Volume);

breakout.pinePine Script

if close > rangeHigh and inLondon

strategy.entry("LB", strategy.long)

strategy.exit("x", "LB", loss=atr)

breakout.rs

Rust

if bar.close > self.high {

let stop = atr / ctx.pip();

ctx.signal(Side::Long, stop, stop * 2.0);

}

breakout.blocks

Blocks

Closeis aboveRange high

Enter longstop1 ATR

A Python engine and a Rust core

The Rust core runs optimisations up to five times faster than the Python engine, and the optimiser keeps every core of your machine busy.

Get early access

Market data and backtests, in one place.

Forex, metals, indices, crypto, equities, futures and CFDs, from minute bars down to every tick, in the workspace you test in. No separate data vendor, no API keys.

7 asset classes
Tick resolution
L2 order book data

Archive at a glance
See which days are there before you download.
Your own data
CSV and plugin sources beside the archive.
GitHub
Link a repository of your own and push the workspace to it.
Three accuracy tiers
OHLC, M1 or tick. The run says which.
Broker costs
Spread, commission and swap on every trade.
No lookahead
Signal on the close, fill on the next bar.

Build on existing research.

Start from what the community publishes, or from a quant paper.

Community
Published strategies
Clone one into your workspace in a click, or publish your own for others to build on.
Clone-ready
Research
SSRN imports
Bring a quant paper in from SSRN and test it as a strategy.
- Time Series Momentum, 2012
- Value and Momentum Everywhere, 2013
Arrives before the alpha release

Plugin library

31 ready made plugins in Python: data sources, indicators, strategy templates and exporters, each with its documentation and its code.

Browse the library

Illustration only, not a result of real trading. OrionQuant is software, not a financial service, and gives no investment advice.

One strategy,three ways to see it.

On the chart

Every trade of the run replayed bar by bar on the chart it was computed from: the range, the entry, the stop and the target, and the quote each fill was taken against.

Orion Workspace

README.md

filters.py

strategy.pynew

Notebook

notes.md

data.md

filters.py

strategy.py

ValidateRun

1from filters import in_session

4class LondonBreakout(Strategy):

5 """Buy a break of the Asian range in London."""

7 def __init__(self, params):

8 self.atr_period = int(params.get("atrPeriod", 14))

9 self.reward = float(params.get("reward", 2.0))

10 self.day = self.high = None

11 self.done = False

13 def on_bar(self, ctx):

14 bar = ctx.bars(1)[-1]

15 if bar.ts.date() != self.day:

16 self.day = bar.ts.date()

17 self.high, self.done = bar.high, False

18 if in_session(bar.ts, "asia"):

19 self.high = max(self.high, bar.high)

20 return

21 if self.done or not in_session(bar.ts, "london"):

22 return

23 atr = ctx.indicators.atr(self.atr_period)

24 if atr and bar.close > self.high:

25 stop = atr / ctx.pip()

26 target = stop * self.reward

27 ctx.signal("LONG", sl_pips=stop, tp_pips=target)

28 self.done = True

Python · 3 filesEURUSD · Tick · Jan to Jun 2024Report ready

London breakout

Build a London breakout for EURUSD. Take the high of the Asian range, buy the first close above it after 07:00 UTC, one ATR stop, 2R target. Use the session helper in filters.py.

Read workspace3 files

filters.py already has in_session, so I will import it rather than repeat it, and write the strategy against the engine's API.

Writing strategy.py+28

+ target = stop * self.reward

+ ctx.signal("LONG", sl_pips=stop, tp_pips=target)

+ self.done = True

Validate

1 strategy class: LondonBreakout
Imports: filters, from this workspace. None refused
Parameters: atrPeriod, reward

Valid, nothing refused. Run it on EURUSD at the tick tier and every fill uses the bid and ask as they were.

Run backtestSaveOptimize

Ask a follow up

Claude Opus 5

In the workspace

The strategy as the file it is, beside the helpers it imports and the notebook where the assistant logs what it did. In Python, C# or Pine Script.

As blocks

The same rule put together from blocks: a value, a comparison, an action. Nothing to type, the code is written from the blocks for you.

Illustration only, not a result of real trading. OrionQuant is software, not a financial service, and gives no investment advice.

Use OrionQuant withClaude.

Connect an AI app on this computer, and Claude Desktop, Claude Code or Cursor reads your strategies and your runs, validates a strategy without running it, and looks up the engine's API instead of guessing at it. Codex and any other client that speaks MCP connect the same way.

It runs on your machine beside the engine, reads a local copy of your library, works offline and costs nothing, and one switch in Settings turns it off. AI apps in the cloud brings your server runs, results and usage to Claude, ChatGPT and other apps in the browser or on your phone.

Claude Code

why did my breakout strategy lose money in March?

I'll read the strategy and its last run in OrionQuant.

orionquant · read_run

BreakoutStrategy · EURUSD · 214 trades · tier TICK

orionquant · validate_strategy

1 strategy class found · no refused import

Most of the loss is eleven entries in the Asian session. I'll add a session filter and validate it again.

One workspace, from idea to verdict.Four tools, one workspace, one subscription.

Analysis

Risk analysis

Monte Carlo, walk forward and an audit before you believe the result.

The run5 to 95 percent of paths

Data

Data manager

See what is available, download it, done.

Builder

Strategy builder

Write it in Python, C# or Pine, put it together from blocks and read it as Rust, or describe it to the assistant.

86 indicators built in, market structure among them: Order blocks, Fair value gaps, Breaker blocks, Liquidity sweep, Fibonacci retracement. Browse the library

Cloud

Server compute

Let it compute on our servers instead of tying up your own machine.

Get early access

Illustration only, not a result of real trading. OrionQuant is software, not a financial service, and gives no investment advice.

Usually

A data vendor for the history
Backtesting software
A server to run it on
An AI subscription

Four tools, four bills, four logins.

OrionQuant

Market data, built in
Engine and optimiser
Server compute
AI assistant and Claude

One workspace, one subscription.

Get early access

Stop paying to fail prop challenges

Audit a backtest and OrionQuant runs ten thousand simulations against the rules of Topstep, FTMO and FundingPips, read from their own pages. Know your chance to pass, to be paid, and what it nets after fees, before you buy an evaluation.

Loading the example run

Illustration only, not a result of real trading. OrionQuant is software, not a financial service, and gives no investment advice.

The whole quant stack.One subscription.

Save your founding price now, while the first fifty seats are open.

Only the first 50 on the waitlist get the founding price.

After that, every plan costs the regular price, the one struck through below. A founding price stays yours for as long as your plan runs, through every upgrade and change of cycle.

Join the waitlist and get a 10 day trial on top.

Free

A real account to learn the engine.

0 €

per month

Free, for as long as you like

Features

3 saved strategies
15 backtests a day, on your own machine
OHLC and minute bar accuracy
5 indicators at a time
1 optimizer run a month
C# strategies and cTrader cBots
50 AI tokens a month

Pro

The whole history, grid search and server runs.

~~39 €~~Save 33%

26 €

per month, incl. VAT

Billed monthly

Founding price for the first 50 on the waitlist.

Everything in Free, and:

Minute bars, the whole history
Grid optimisation, in and out of sample
25 strategies and indicators
300 server minutes a month
5 audits a month
CSV and report export
12,000 AI tokens a month

Max

Enterprise

A desk of seats, one contract, your sign in.

On request

plus VAT

Invoiced annually

Talk to sales

Everything in Max, and:

Five seats, more on request
Roles within the team
Single sign on over OIDC
No limits on runs, audits or AI tokens
Priority support

Cancel anytimeSecure checkout via StripeNo hidden feesPrices in euro incl. VAT. Enterprise is for businesses, plus VAT.

The market alreadytrades this way.

Share of volume traded by algorithm

75%Forex
70%Futures
65%Crypto

Estimates. Sources: BIS 2022, CFTC, industry 2025

Most of the volume is already traded by rules. Test yours the same way: on institutional data, stressed with Monte Carlo, with the costs a broker charges.

From zero code to a backtest at full speed.

Market data and backtests, in one place.

Archive at a glance

Your own data

GitHub

Three accuracy tiers

Broker costs

No lookahead

Build on existing research.

Published strategies

SSRN imports

Plugin library

One strategy,three ways to see it.

Use OrionQuant withClaude.

One workspace, from idea to verdict.Four tools, one workspace, one subscription.

Risk analysis

Data manager

Strategy builder

Server compute

Stop paying to fail prop challenges

The whole quant stack.One subscription.

Free

Pro

Max

Enterprise

The market alreadytrades this way.