Skip to main content
Agents are autonomous programs that execute trading strategies on your behalf. Each agent runs on a configurable schedule, analyzes market conditions using The Brain, and executes trades on its own dedicated Hyperliquid wallet.

​
Overview

An agent consists of five components:

Instructions

The rules that define when and how your agent trades

Goal

Your strategic objective and risk parameters

Trigger

The schedule or signal that tells your agent when to run

Memory

Learning edge that tracks past trades and patterns to improve future decisions

Wallet

A dedicated Hyperliquid wallet for the agent to trade with

​
How agents work

​
Execution cycle

Every time your agent’s trigger fires, it follows this sequence:
  1. Query current state: Retrieve wallet balance, open positions, and recent performance
  2. Consult memory: Review past trades under similar conditions from the Learnings database
  3. Analyze market conditions: Pull macro, microstructure, price action, fundamentals, and sentiment data as needed
  4. Evaluate instructions: Determine whether entry, exit, or risk management criteria are satisfied, informed by both current analysis and historical outcomes
  5. Execute actions: Place orders, modify positions, or do nothing based on evaluation
  6. Record outcome: Log the reasoning, actions taken, and market state to the Learnings database for future reference

​
Dedicated wallets

Each agent gets its own Hyperliquid wallet. This keeps your agents isolated from each other and from your personal funds. The flow is:
  1. Create your agent
  2. Generate a wallet for the agent
  3. Fund the wallet with USDC
  4. Activate the wallet
  5. Enable the agent
Always export your agent’s wallet private keys before deleting an agent. Deletion is permanent and any remaining funds in the wallet are irrecoverable. See the Agent Setup Guide for details.

​
Memory and decision making

Unlike traditional bots that treat every market condition the same way, Gigabrain Agents build institutional knowledge from every trade. Before executing, your Agent reviews past trades under similar conditions, checking what worked, what didn’t, and why. What gets recorded:
  • Trade outcomes relative to initial analysis
  • Market conditions when trades succeeded or failed
  • Patterns in profitable vs. unprofitable setups
  • Strategy adjustments made over time
How memory improves decisions: Before executing any trade, Agents consult their memory to review past performance under similar market conditions. For example, if your Agent took three losing trades during low-volume consolidation periods despite meeting all entry criteria, it will recognize this pattern and apply more caution when similar conditions appear again, potentially requiring stronger confirmation signals or skipping marginal setups entirely. This memory doesn’t automatically modify your Instructions, but it informs how the Agent interprets market conditions and applies your rules. Over time, this creates institutional knowledge specific to your strategy. The Agent develops “track record awareness” that improves execution quality without requiring manual backtesting or constant instruction updates.

​
Configuration

​
Instructions

Define your trading logic using natural language. Be explicit about entry criteria, position sizing, exit rules, and risk limits. 
Monitor BTC and ETH on the 15m timeframe.

ENTRY CRITERIA:

- Price breaks above previous 4H high with volume >150% of 20-period average
- RSI(14) between 55-70 (momentum confirmed, not overbought)
- Funding rate <0.03% (avoiding overleveraged longs)
- Taker buy volume >55% over last 3 candles (aggressive buying)

FILTERS (DO NOT ENTER IF):

- VIX >30 (macro risk-off environment)
- Major resistance within 2% above entry
- Ongoing high-impact news event within next 2 hours

POSITION SIZING:

- Base size: 5% of account balance
- Leverage: 3x (15% notional exposure)
- Maximum 2 concurrent positions

EXIT RULES:

- Take profit: +6% from entry (trim 50%), +10% (close remaining)
- Stop loss: -3% from entry (hard stop)
- Time stop: Close after 48 hours if neither TP nor SL hit
- Trailing stop: Once +5%, trail stop at -2% from peak

RISK MANAGEMENT:

- Daily loss limit: -5% of account (halt all trading)
- Weekly profit target: +15% (reduce position sizes by 50%)
- Maximum 4 trades per day (prevent overtrading)
Instructions are interpreted by the agent’s language model. Ambiguous phrasing may lead to unexpected behavior. Test new instructions with small position sizes.

​
Triggers

Choose how your agent decides when to run:
TypeHow it worksBest for
Every 5 minutesRuns on a fixed scheduleScalping strategies
Every 15 minutesRuns on a fixed scheduleIntraday momentum trading
HourlyRuns on a fixed scheduleSwing trading setups
Every 4 hoursRuns on a fixed schedulePosition trading
DailyRuns on a fixed scheduleFundamental-based strategies
Alpha triggerRuns when a matching signal is detected by the Alpha engineEvent-driven and signal-based strategies
Alpha triggers use vector matching against incoming Alpha signals. The more specific your match string, the better the filtering. You can also set a minimum impact rating (1-5) to only trigger on high-conviction signals.

​
Monitoring

Each agent run generates a trace showing:
  • Current wallet state (balance, open positions, P&L)
  • Market analysis across relevant domains
  • Evaluation of entry/exit criteria
  • Actions taken or reasons for inaction
Use this to verify the agent interprets your instructions correctly.

​
Example strategies

Objective: Identify and ride explosive moves across any asset.
  • Trigger: Every 15 minutes
  • Analysis: Market State, Trenches, Price Movement
  • Entry: High-conviction Alpha signal (>75% confidence)
  • Position sizing: 5% of wallet per trade, 5x leverage
  • Exit: Trailing stop at 2% below peak, manual review after 24h
  • Risk limit: Maximum 3 concurrent positions
Objective: Take directional positions when technical and macro factors align.
  • Trigger: Every 4 hours
  • Analysis: Macro, Market State, Price Movement
  • Entry: 4H chart setup + favorable risk regime + positive BTC correlation
  • Position sizing: 10% of wallet, 3x leverage
  • Exit: TP at +8%, SL at -4%
  • Risk limit: No new entries during risk-off regimes
Objective: Trade based on on-chain metrics and protocol health.
  • Trigger: Daily at 00:00 UTC
  • Analysis: Fundamentals, Market State
  • Entry: Daily protocol fees >$3M, TVL increasing week-over-week
  • Position sizing: 10% of wallet, 2x leverage
  • Exit: Hold until fundamentals deteriorate or -10% SL
  • Risk limit: Maximum 2 protocol-based positions

​
Best practices

Start conservative

Use small position sizes (1-5%), set tight stops, and monitor the first 10-20 reasoning traces closely before scaling up.

Iterate based on learnings

Review the Learnings tab weekly to identify failure modes, optimal conditions, and opportunities to refine entry criteria.

Use daily caps

Set profit caps at 2-5% of wallet value to prevent overtrading and lock in gains.

Monitor macro regime

Reduce position sizes during high VIX, avoid mean-reversion in trending Fed policy, pause during major geopolitical events.

Export keys regularly

Back up your agent wallet’s private keys. If you ever need to delete an agent, you’ll need the keys to recover funds.

Review reasoning traces

Regularly check traces to verify correct interpretation and identify edge cases in your strategy logic.

​
Troubleshooting

Possible causes:
  1. Agent is not enabled (check agent state is “Online”)
  2. Wallet is not activated or funded
  3. Entry criteria are not being met (review reasoning traces)
  4. Daily cap has been reached (check recent P&L)
  5. Risk limit has been triggered (check wallet balance vs. floor)
Possible causes:
  1. Instructions are ambiguous (refine phrasing and test again)
  2. Market conditions changed rapidly between trigger intervals
  3. Market data was stale or delayed
  4. Edge case not covered in your instructions
Review the reasoning trace for the specific run to understand what triggered the trade.
Possible causes:
  1. Market regime has shifted (e.g., trending to range-bound)
  2. Strategy has attracted attention (edge degradation)
  3. Position sizing is too large for current volatility
  4. Stop losses are too tight for current ATR
Review learnings for patterns and adjust instructions accordingly.

​
Limitations

Agent actions are subject to trigger interval (minimum 5 minutes), API latency to Hyperliquid, order processing time, and network conditions. Do not use agents for strategies requiring sub-minute execution.
Agents cannot predict unexpected events (exchange outages, liquidity crises), guarantee fills at expected prices during volatile markets, or prevent slippage on large orders in illiquid markets.
The language model may misinterpret ambiguous phrasing, cannot execute logic requiring external data not available through its tools, and has a knowledge cutoff. Review reasoning traces to verify correct interpretation.

​
Next steps

Agent Setup Guide

Step-by-step instructions for creating your first agent

Agent Builder Reference

Strategy archetypes and design patterns
Not financial advice: Agents execute your strategy, but they do not provide financial advice. You are responsible for defining appropriate risk parameters, understanding the strategies you implement, monitoring agent performance, and complying with relevant regulations.