Documentation

Everything you need to run it,
in the order you need it

Start with the quickstart, which gets a strategy backtesting through the API in about ten minutes. The concepts section explains the vocabulary the rest of the platform uses, and the guides cover the first things people set up.

Quickstart

Install the Python client, create a key, and run a backtest. This does not touch an exchange and cannot place an order.

1 · install

pip install "git+https://github.com/Arjuunb/nexus-trading-bot#subdirectory=sdks/python"

2 · authenticate

export NEXUS_API_KEY="nxs_..."   # Settings → Security → API keys (control scope to queue backtests)

3 · backtest

from tradelogx_nexus import Client

client = Client()               # reads NEXUS_API_KEY

job = client.backtests.run(
    "brain",                    # an id from client.strategies.list()
    symbol="BTCUSDT",
    timeframe="15m",
    bars=1000,
)

gross, net = job["result"]["gross"], job["result"]["net"]
print(gross["expectancy_r"], net["expectancy_r"])   # before and after costs
print(job["result"]["costs"]["net_r_drag"])         # what the costs took, in R

Before you trust a backtest

A backtest that looks good is a hypothesis, not a result. Run the strategy in paper mode against the live feed for a full month and compare — it costs nothing, and live routing stays locked until a strategy has earned it. The risk disclosure covers why this matters more than it sounds.

Guides

Task-shaped, not feature-shaped. Each one ends with something working.

5 min

Attach an exchange key

In Settings → Security, paste a Binance key. Binance is asked what the key may do first: a key that can withdraw or transfer funds is refused before it is stored. Bind it to the server's IP on Binance. Live routing is locked, so the key is kept encrypted for later.

10 min

Run your first backtest

Queue one through the API or run it in the dashboard's Backtesting page. Results come back twice — gross, and net of modelled fees, spread and slippage — so a strategy whose edge is only its costs shows as exactly that.

8 min

Set your risk limits

Daily and weekly loss limits, per-position and total exposure caps, correlated-position limits, session windows and a drawdown halt, in Settings. An order that fails any check is never created.

3 min

Paper trading and the live lock

Every strategy runs in paper mode against the live Binance feed, with modelled costs and a full journal. Live order routing is locked; asking the API to promote a strategy to live answers live_routing_locked.

Concepts

The vocabulary the API, the dashboard and the journal all assume. Worth ten minutes before the reference.

Quality score: The Decision Brain's 0–100 score from eight weighted factors — higher-timeframe alignment, regime, reward : risk, momentum, stop safety, volatility, structure and volume. 60 is the default minimum; some setups are blocked whatever they score.
Regime: Trending, ranging, or high, low or extreme volatility. Regime fit is part of the score, and a choppy regime blocks trades that are not reversals.
Decision: Every evaluation — accepted or rejected — is stored with its scores, the rules it passed and failed, and the reason. The full market inputs are not stored yet, so a decision cannot be re-run.
Risk checks: Twenty checks on the one path from a signal to an order: loss limits, exposure, correlation, event blackouts, sessions, venue lot rules and more. If a check cannot run, nothing trades.
Trade memory: Every closed trade with its context, searchable by similarity. It is not consulted when a decision is made; repeated losing patterns reduce trade size instead.
Order intent: A paper order waiting for a Binance quote to fill it. A limit entry that is never reached expires rather than being chased.

Where to go next

The reference is exhaustive; these three pages are the ones people need first.

API referenceEvery endpoint, with request and response shapes SDKsPython, TypeScript, Go and Rust clients How it worksThe seven stages a trade passes through

Something missing or wrong here is a documentation bug and worth reporting the same way as any other — through the support center.

Quickstart#

Guides#