Deploy

This is the moment your algorithm moves from simulation to the real market. The live engine uses the exact same state machine and logic as the backtester — what you tested is what runs. No surprises from implementation differences.

For a hands-on first deployment, see the Deploy Walkthrough →

On this page: Going Live · Monitoring · Run States · Stopping and Restarting · Tuning a Live Strategy

Note

Going live is exciting — and a little nerve-wracking. Start small, monitor closely, and scale up once live performance matches your backtest.

---

Going Live

Before You Deploy

Make sure you’ve done these before switching to live:

• Backtested on at least 6 months of data
• Checked that your Sharpe ratio is above 1.0 and max drawdown is acceptable
• Ideally run rolling cross-validation to confirm the strategy generalizes
• Decided on a conservative Starting Capital to launch with

Deploy from Run Setup

Switch to the Run tab and open Run Setup. Pick your wallet, your algorithm, the trading symbol, your Starting Capital and leverage, then click Start Run. The algorithm begins executing in real-time using the exact same engine the backtester uses — what you tested is what runs.

Starting Capital and Leverage

The Run Setup panel asks for two numbers that together drive position sizing:

effective notional = Starting Capital × leverage

Starting Capital	Good For
$50 – $100	First deployment — minimal real exposure while you verify behavior
$250 – $500	Growing confidence — you’ve watched it live for a few days
$1,000+	Full deployment — live performance consistently matches backtest

Leverage in the live run can match what you backtested; it sizes positions relative to the Starting Capital, not your wallet balance.

Caution

Even a well-backtested strategy can behave differently with real slippage and latency. Start with a small Starting Capital, stop the run after a few days once you’ve confirmed live execution matches expectations, and relaunch with a larger Starting Capital.

---

Monitoring

The First 24 Hours

Pay close attention during the first day of a live deployment. Here’s what to watch:

• Fill prices — are entries and exits landing close to where the backtest expected them? Some slippage is normal, but consistent deviation may mean your order type or timeframe needs adjusting.
• Trade frequency — is the algorithm trading as often as the backtest predicted? If it’s much more or less active, market conditions may have shifted.
• Drawdown — if your live drawdown exceeds your backtest’s max drawdown early, consider pausing to investigate.

Note

Compare your first few live trades against what the backtest showed for the same period. If the fills, timing, and P&L are in the same ballpark, you’re on track.

Run Mode Dashboard

Run mode gives you dedicated monitoring panels:

• Run Monitor — real-time view of all active algorithms with position, entry price, last price, equity, trade count, unrealized P&L, and equity sparkline
• Run History — completed and stopped runs with performance summary
• Run Chart — live chart visualization for the active run
• Configuration — the algorithm settings for the current run
• Active Positions tab (Info panel) — open positions with real-time P&L
• Trade History tab (Info panel) — every entry and exit with P&L

Via Agent Tools

Use the MCP agent tools for programmatic monitoring:

Tool	What It Shows
list_runs	All runs (backtests and live) with status, filtered by algoId or status
get_backtest_results	Metrics for a completed run
get_account_overview	Account snapshot including active positions and running algos

---

Run States

State	What It Means
NEW	Run created, initializing
RUNNING	Actively executing — watching the market and placing trades
STOPPING	Stop requested, winding down gracefully
STOPPED	You (or the system) paused the algorithm. Open positions remain unless you close them
DONE	Run completed normally
FAILED	Run encountered an error
CANCELLED	The run was aborted before completion

You’ll typically see RUNNING after deployment. If something looks off, you can stop it at any time.

---

Stopping and Restarting

Stop a Run

Set the run status to STOPPED from Run mode or via the agent:

"Stop my running algorithm"

Or use the update_run_status tool: update_run_status(runId, "STOPPED")

Caution

Open positions are not automatically closed when you stop an algorithm. Close them manually from Trade mode or via agent tools (close_position). Always check your positions after stopping a run.

Iterate and Redeploy

1. Stop the current run
2. Switch to Build mode and edit the algorithm (creates a new version)
3. Switch to Test mode and backtest the new version to validate changes
4. Switch to Run mode and deploy the updated version

Note

Each edit creates a new version, so you can always compare performance across versions. You can even run multiple versions simultaneously to A/B test changes.

---

Tuning a Live Strategy

Signs Your Strategy Needs Tuning

• Live Sharpe is significantly below backtest Sharpe — the strategy may be overfit to historical data
• Drawdowns are larger than expected — risk parameters may need tightening
• Trade frequency has dropped off — the market regime may have shifted away from your strategy’s edge
• Most exits are stop losses — entry timing or stop placement may need work

What to Adjust

Based on live performance, go back and tweak:

• Thresholds — if signals trigger too frequently or rarely
• Indicators — swap underperforming indicators (check usefulness scores from backtests)
• Position sizing — reduce if drawdowns are too large
• Stop loss — tighten or loosen; experiment with trailing stops

Re-optimization

If the market regime shifts, re-run optimization and cross-validation on recent data:

1. Run a backtest on recent data to see current performance
2. Re-optimize if metrics have degraded
3. Validate with rolling cross-validation
4. Deploy the new optimized version

See Backtest → Optimization for the full optimization workflow.

---

What’s Next?

Want to…	Go to
Set up autonomous AI trading	Agent Features →
Re-optimize your strategy	Backtest Features →
Walk through your first deployment	Deploy Walkthrough →