Skip to main content

Creating And Configuring An Agent

This walkthrough covers the full lifecycle of an Intelligence Agent: create, edit, tune, and run.

Step 1: Create Your Agent​

The Create Agent page is intentionally scoped to identity + strategy only. Trading configuration is added later on the edit page. This keeps the onboarding contract simple.

On the create page you configure:

  1. Identity Strip — agent name, short bio, and avatar.
  2. Avatar — open the picker to choose a curated robot avatar from the catalog, or hit randomize to regenerate a DiceBear seed.
  3. Strategy Build Path — pick a curated strategy preset or fork to Custom for a fully editable strategy prompt.
  4. Context Sources — if you're on the Custom path, toggle the market-data modules the agent can read (e.g. candles, correlations, regime indicators).
  5. Live Context Preview — preview any enabled module in real time for a selected game type and timeframe.

When the form is complete, save. Your new agent lands in OFF mode by default and is immediately available on the edit page.

tip

Curated strategy presets are a fast way to get a sensible baseline. You can always fork a preset into Custom later to tweak individual behaviors.

Step 2: Open The Edit Page​

The edit page is organized into five tabs:

TabWhat It Owns
StrategyBehavior, context sources, overlay text, strategy preset/custom fork
AutomationRuntime controls (informational — owns no persisted form fields)
SignalsSignal aggressiveness preset, score threshold, ATR floor, stop-loss range, risk/reward, timeout, entry deviation
TradesTrading mode (OFF / SIGNALS_ONLY / FULL_EXECUTION), account balance limits, leverage, allocation, position sizing
PositionBreak-even, trailing stop, time-decay, take-profit ladder

Each tab (except Automation) shows a dot indicator beside the label whenever that slice has unsaved changes. One Save commits everything at once.

Step 3: Configure Strategy​

The Strategy tab mirrors the create flow:

  • Preset path — pick a named strategy from the browser. Read-only detail sheets let you inspect the full prompt, behavior, and context source list before committing.
  • Custom path — edit your own strategy prompt, toggle context sources, and adjust overlay text.

Switching between preset and custom forks is explicit. If you jump from a Preset back to Custom, a confirmation dialog warns you before replacing your custom state.

Step 4: Configure Signals​

The Signals tab is where you tune the agent's entry aggressiveness. Two paths:

  • Pick a named preset: FOXHOLE, FRONTLINE, or BLITZKRIEG. Each preset sets all eight signal-aggressiveness fields at once.
  • Fork to CUSTOM: edit any individual field. As soon as you touch a signal field on a named preset, the preset automatically flips to CUSTOM so you know you've diverged.

The 8 signal fields are:

FieldMeaning
Composite score thresholdMinimum weighted signal score required to open a trade
Primary signal countMinimum number of top-tier signals required
ATR floorMinimum volatility required (avoids dead markets)
Stop-loss range (min/max)Allowed SL distance bounds
Risk/reward floorMinimum R:R ratio (0.5–5.0, default 1.5)
Signal timeoutSIGNALS_ONLY approval window (5 / 10 / 15 min)
Max entry deviation ATR multiplePrice-drift invalidation for pending entries
tip

Presets never overwrite capital or execution settings. Leverage, wallet limits, and position management stay exactly where you left them.

Step 5: Configure Trades​

The Trades tab owns capital and execution ownership:

  • Trading mode — OFF / SIGNALS_ONLY / FULL_EXECUTION.
  • Account balance threshold — minimum wallet balance required before the agent will attempt a trade.
  • Leverage — actively enforced on the exchange before each order.
  • Allocation percent — how much of available capital each trade is allowed to use.
  • Position size preset — SMALL / MEDIUM / LARGE. Defined as percentages (MANUAL mode) or auto-calibrated from ATR (VOLATILITY_AUTO mode).
  • Max daily trades — daily rate limit.
  • Minimum order guard — respects HyperLiquid's $10 exchange minimum.

Field changes in this tab never flip the Signals preset.

Step 6: Configure Position Management​

The Position tab is opt-in per feature. Flip the master toggle to enable the subsystem, then enable each feature you want:

  1. Break-Even Protection — set the ROE trigger and an optional offset from entry.
  2. Trailing Stop — pick fixed percent or ATR-based, set the value, and choose when to activate (immediately, after break-even, or after the first TP hit).
  3. Time-Decay Stop — set the grace period, interval, per-step tighten percent, max tighten cap, and stale-ROE threshold.
  4. Take-Profit Ladder — add up to 10 levels. Each level closes a percent of the remaining position when ROE crosses a trigger. Triggers must be strictly increasing and total close-percents must stay under 100.

See Position Management for deeper guidance on each feature.

Step 7: Save And Run​

Hit Save. The server validates your configuration, bumps the agent's revision number, and returns the updated record. From there:

  • If the agent is in FULL_EXECUTION, it starts evaluating signals on the next pipeline cycle.
  • If it's in SIGNALS_ONLY, you'll see toast notifications when a signal arrives, with Accept / Deny buttons that respect your configured timeout window.
  • If it's in OFF, it's idle until you flip the mode.

Editing A Running Agent​

  • Concurrent edits: if someone (or another tab) saved the agent between the time you loaded it and now, the server rejects the save with a 409 Conflict. The UI prompts you to refresh and reapply your changes.
  • Inline overlay edits: quick edits from the in-game chat overlay reuse the same save contract. You can tweak an agent mid-session without losing concurrency protection.
  • Unsaved-changes indicators: dot badges appear on Strategy, Signals, Trades, and Position whenever their local draft diverges from the server copy. Automation has no badge because it owns no persisted fields.

Deleting Or Pausing​

  • Pause: set Trading Mode to OFF. The agent stops evaluating immediately. Existing positions are not closed.
  • Delete: removes the agent and its configuration. Historical signal and decision records remain for audit purposes.
tip

Always pause before major strategy rewrites. Changing the strategy prompt while the agent is in FULL_EXECUTION can produce surprising signals in the next cycle.

Next: explore Trading Configuration for preset details or Position Management for protection features.