Building AI Agents & Strategies
Master agent configuration, strategy customization, and building specialized trading agents for your goals.
While Agentical provides powerful pre-built trading strategies, the platform's true versatility emerges when you customize agents to match your specific trading philosophy, risk tolerance, and market perspective. This page guides you through the process of building personalized AI trading agents that execute your unique strategy with mechanical precision.
Understanding Agent Architecture
What Makes an Agent
An AI trading agent in Agentical is more than simple automation—it's a configurable intelligence system that interprets market data through your strategic lens and executes decisions according to your rules.
Every agent consists of three fundamental components working in harmony:
Strategy Configuration defines what the agent looks for in potential trades. This includes specific criteria tokens must meet, risk thresholds that cannot be exceeded, and market conditions that favor entry or exit. Think of this as the agent's instruction manual—the rulebook it follows without deviation.
Execution Parameters control how the agent behaves when opportunities arise. This encompasses position sizing logic, entry timing preferences, budget constraints, and portfolio management rules. These parameters transform strategy into action.
Learning Preferences determine how the agent adapts over time. You can configure whether the agent should automatically refine criteria based on performance, maintain strict adherence to original settings, or operate in a hybrid mode where certain parameters evolve while others remain fixed.
Agent Component Architecture
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
┌─────────────────────────────────────────┐
│ STRATEGY CONFIGURATION │
│ ┌─────────────────────────────────┐ │
│ │ • Token Criteria & Requirements │ │
│ │ • Risk Tolerance Thresholds │ │
│ │ • Market Condition Filters │ │
│ │ • Entry/Exit Trigger Rules │ │
│ └─────────────────────────────────┘ │
└───────────────┬─────────────────────────┘
↓
┌─────────────────────────────────────────┐
│ EXECUTION PARAMETERS │
│ ┌─────────────────────────────────┐ │
│ │ • Position Sizing Logic │ │
│ │ • Budget & Limit Controls │ │
│ │ • Timing & Entry Optimization │ │
│ │ • Portfolio Management Rules │ │
│ └─────────────────────────────────┘ │
└───────────────┬─────────────────────────┘
↓
┌─────────────────────────────────────────┐
│ LEARNING PREFERENCES │
│ ┌─────────────────────────────────┐ │
│ │ • Performance-Based Adaptation │ │
│ │ • Criteria Refinement Settings │ │
│ │ • Static vs Dynamic Parameters │ │
│ └─────────────────────────────────┘ │
└─────────────────────────────────────────┘Agent Types & Templates
Agentical provides three foundation templates, each representing a distinct trading philosophy. Understanding these templates helps you choose the right starting point for customization.
Conservative Agent Template
The conservative template prioritizes capital preservation and consistency over aggressive profit seeking. This agent operates with strict entry requirements, smaller position sizes, and tight risk controls.
Philosophy: Generate steady returns through high-probability setups while minimizing drawdown risk.
Typical Characteristics:
Risk score requirement: 75+ (vs 65+ balanced)
Holder count minimum: 1,500+ (vs 1,000+ balanced)
Token age window: 8-48 hours (avoids brand new launches)
Top holder concentration: Maximum 3% (vs 5% balanced)
Profit targets: 25-35% (realistic, achievable)
Stop losses: -10% (tight protection)
Position sizing: 60% of standard allocations
Balanced Agent Template
The balanced template strikes equilibrium between growth and protection, suitable for users seeking moderate risk-adjusted returns.
Philosophy: Capture quality opportunities across various market conditions while maintaining reasonable risk management.
Typical Characteristics:
Risk score requirement: 65+
Holder count minimum: 1,000+
Token age window: 4-48 hours
Top holder concentration: Maximum 5%
Profit targets: 40-60%
Stop losses: -15%
Position sizing: Standard allocations
Aggressive Agent Template
The aggressive template targets high-growth opportunities with relaxed entry criteria and larger position sizes, accepting increased volatility for potential outsized returns.
Philosophy: Capture explosive early-stage growth by entering emerging opportunities before they mature.
Typical Characteristics:
Risk score requirement: 55+ (vs 65+ balanced)
Holder count minimum: 750+ (vs 1,000+ balanced)
Token age window: 2-72 hours (wider window)
Top holder concentration: Maximum 7% (vs 5% balanced)
Profit targets: 60-100% (ambitious)
Stop losses: -20% (wider tolerance)
Position sizing: 120% of standard allocations
The Agent Building Process
Step 1: Define Your Trading Philosophy
Before configuring technical parameters, clarify your fundamental trading approach and constraints.
Essential Questions to Answer:
Risk Tolerance Assessment - How much volatility can you emotionally handle? Can you tolerate seeing 20-30% drawdowns during losing streaks, or do losses exceeding 10% cause significant stress?
Time Horizon Expectations - Are you seeking quick scalps measured in minutes and hours, or swing trades that develop over days? Your answer fundamentally shapes agent configuration.
Capital Allocation Strategy - What percentage of your overall portfolio will this agent manage? Are you dedicating speculative capital you can afford to lose, or is this more conservative capital requiring strict protection?
Performance Goals - What defines success for this agent? Is it absolute returns, risk-adjusted returns, win rate consistency, or learning opportunities about market dynamics?
Monitoring Commitment - How actively will you monitor the agent? Daily check-ins, weekly reviews, or mostly passive with occasional oversight?
These answers create the philosophical foundation upon which you'll build specific agent parameters.
Step 2: Choose Your Starting Template
Based on your trading philosophy, select the template that most closely aligns with your approach.
Template Selection Decision Tree:
Remember that templates are starting points—you'll customize from here to create your unique strategy.
Step 3: Customize Core Criteria
With a template selected, begin customizing the criteria that define which tokens your agent will trade.
Criteria Customization Interface:
The platform provides both guided interfaces and natural language instruction capabilities for criteria customization.
Guided Interface Method:
Each criterion you add makes the agent more selective. More criteria doesn't necessarily mean better performance—it means fewer trades with (hopefully) higher quality.
Natural Language Method:
Advanced users can provide criteria using conversational instructions:
Example Natural Language Instructions:
The AI interprets your instructions, validates feasibility, warns about potential issues, and seeks clarification when needed.
Step 4: Configure Risk Parameters
Risk parameters define how the agent manages capital and protects against losses.
Risk Configuration Panel:
Understanding Position Sizing Strategies:
Fixed Size provides consistency and predictability. Every trade receives the same capital allocation regardless of perceived opportunity quality. This simplifies tracking and ensures uniform risk exposure.
Percentage Based automatically scales with your account balance. As profits accumulate, position sizes grow proportionally. During drawdowns, positions shrink, providing natural risk reduction when performance suffers.
Risk-Adjusted allocates capital dynamically based on opportunity quality. Higher risk scores receive larger positions, while marginal opportunities get smaller allocations. This approach maximizes capital efficiency but creates less predictable position sizing.
Step 5: Define Entry & Exit Logic
Entry and exit logic determines when the agent acts on opportunities that meet your criteria.
Entry Logic Configuration:
Your agent might find tokens meeting all criteria, but timing still matters significantly. Entry logic controls when the agent actually executes trades.
Entry Timing Strategies:
Exit Logic Configuration:
Exit strategy often determines profitability more than entry quality. Configure multiple exit triggers to handle different scenarios.
Multi-Trigger Exit System:
The scaled exit strategy often performs better than single-target exits. Taking partial profits at earlier targets locks in gains while allowing remaining position to capture extended moves.
Step 6: Set Budget & Operational Constraints
Financial guardrails ensure your agent operates within acceptable boundaries.
Budget Configuration:
Conservative budget limits protect against unexpected issues. If testing a new strategy, start with smaller per-trade allocations until performance validates the approach.
Step 7: Configure Learning & Adaptation
Determine how your agent evolves over time based on performance feedback.
Learning Configuration:
Learning Mode Comparison:
Static Mode maintains your original configuration permanently. Choose this when testing specific hypotheses or when you want complete control over all parameters. Performance data is collected but not applied automatically.
Guided Learning allows minor optimizations within boundaries you define. The agent might adjust entry timing by a few percentage points or refine profit targets based on success patterns, but won't fundamentally alter strategy without your approval.
Autonomous Learning grants full optimization authority to the AI. It continuously refines all parameters based on performance data, potentially transforming strategy significantly over weeks or months. This mode requires substantial trust and sufficient capital to withstand experimental adjustments.
Advanced Agent Customization
Creating Specialized Agent Strategies
Beyond modifying templates, you can build highly specialized agents targeting specific market niches or trading styles.
Example: "Migration Hunter" Agent
This specialized agent focuses exclusively on tokens approaching or completing pump.fun migration to DEX platforms.
Strategy Configuration:
Example: "Degen Scalper" Agent
This ultra-short-timeframe agent targets quick profits from volatile moves.
Strategy Configuration:
Example: "Blue Chip Hunter" Agent
This conservative agent seeks only the highest-quality emerging projects.
Strategy Configuration:
Multi-Agent Portfolio Strategies
Users with higher AGNT holdings can deploy multiple agents simultaneously, each targeting different market segments.
Example: Three-Agent Portfolio
This approach captures opportunities across market segments while managing overall portfolio risk through diversification.
Testing & Validating Your Agent
The Criteria Testing Tool
Before activating a new agent configuration, use the criteria testing tool to validate its parameters against historical data.
Testing Interface:
This testing reveals whether your criteria are too strict (few matches), too loose (many poor matches), or appropriately balanced for your goals.
Backtesting Custom Strategies
The platform can backtest your complete agent configuration against historical market data.
Backtest Configuration:
Backtesting provides valuable insights but remember: past performance doesn't guarantee future results. Markets evolve, and what worked historically may not work identically going forward.
Agent Management Best Practices
Iterative Refinement Process
Building the perfect agent is an iterative process, not a one-time configuration.
Recommended Refinement Cycle:
Small, measured adjustments allow you to understand what actually improves performance versus random variance.
Common Configuration Mistakes
Avoid these frequent errors when building custom agents:
Over-Optimization occurs when you add too many criteria, making the agent so selective it rarely trades. If your agent hasn't executed a trade in 48+ hours despite active markets, criteria are likely too restrictive.
Conflicting Logic happens when requirements contradict each other. For example, wanting "early entry" (2-hour-old tokens) while requiring "strong holder base" (typically needs 12+ hours to develop) creates impossible conditions.
Unrealistic Expectations manifest as profit targets of 100%+ with stop-losses of -5%. Early-stage token trading requires balanced risk-reward ratios. Expecting huge gains while accepting minimal risk leads to frustrated abandonment before the strategy can prove itself.
Excessive Complexity emerges when agents have 20+ criteria with intricate conditional logic. Start simple, add complexity only when needed. Complex strategies are harder to understand, debug, and optimize.
Saving & Sharing Strategies
Strategy Library Management
Agentical allows you to save multiple agent configurations and switch between them based on market conditions.
Strategy Library Interface:
This library enables rapid switching between proven configurations as market conditions change, without losing your carefully developed strategies.
Community Strategy Sharing
Advanced users can share successful strategies with the community, while others can learn from proven configurations.
Community Strategies:
Top-performing community strategies appear in a shared library where users can view configurations, performance stats, and implementation notes. You can clone community strategies as starting points for your own customization while maintaining your unique tweaks private.
The Art of Agent Building
Building effective AI trading agents combines technical configuration with strategic thinking and patience. The most successful agents aren't necessarily the most complex—they're the ones aligned with realistic goals, properly tested, and given time to prove themselves.
Start with pre-built templates, make small iterative improvements based on data, maintain disciplined risk management, and remember that even the best-configured agent will have losing periods. What matters is consistent edge over time, not perfection in every trade.
Your custom agent represents your unique market perspective translated into automated execution. When properly configured and patiently refined, it becomes a tireless trading partner executing your strategy with unwavering discipline while you focus on higher-level strategic decisions.
Last updated