Creating Custom Agents

Build Your Own AI Specialists

One of CEO.ai's most powerful features is the ability to create custom agents tailored to specific skills, domains, or methodologies. Whether you need a specialized architect for microservices projects or an executor expert in React testing patterns, you can build it—and potentially earn passive income when the CEO Agent selects your creation for projects.

Why Create Custom Agents?

Specialization

Build agents focused on specific frameworks, languages, or methodologies

Consistency

Ensure your preferred coding standards and patterns are followed

Earning Potential

When the CEO Agent selects your agents, you earn credits

Team Extension

Create agents that embody your team's expertise and best practices

Creating Your First Agent

Navigate to Agents

Access the Agents section:

• From your Dashboard, click the Agents card
• Or go directly to the Agents page via the main navigation

Initiate Agent Creation

Click the Create New Agent button to open the agent creation form.

Configure Basic Information

Agent Name

Choose a clear, descriptive name that indicates the agent's specialty.

Examples:

• "React TypeScript Expert"
• "Python FastAPI Architect"
• "Database Schema Specialist"
• "Frontend Testing Guru"

Tips:

✓ Be specific about the agent's domain
✓ Keep it professional and searchable
✗ Avoid generic names like "Helper Agent"

Description

Write a clear description of what your agent does and when it should be used.

Good Example:

"Specialized in architecting React applications with TypeScript, focusing on component composition, custom hooks, and state management patterns using Zustand. Follows functional programming principles and emphasizes type safety."

What to include:

• Primary technology/domain
• Specific expertise areas
• Methodology or approach
• Best use cases

Model Selection

Choose the AI model that powers your agent from the dropdown menu.

Default: Claude Sonnet 4.5

• Excellent balance of capability and cost
• Strong coding abilities
• Good reasoning and problem-solving

Agent Type

Select whether your agent is an Architect or Executor:

Architect

• Designs overall project structure
• Creates comprehensive task lists
• Makes high-level architectural decisions

Executor

• Implements specific, well-defined tasks
• Writes code, configurations, or documentation
• Follows architectural specifications

Important: Regardless of type, you can have direct conversations with your agent for any purpose. The type primarily affects how the CEO Agent uses your agent in MultiTask projects.

Define the System Prompt

The system prompt is the foundation of your agent's identity and capabilities.

What Is a System Prompt?

The system prompt tells the AI model:

• What role it's playing
• What it's good at
• How it should approach problems
• What standards or principles to follow

System Prompt Example:

You are a senior React developer specializing in building scalable, type-safe applications using TypeScript and modern React patterns.    
    
Your expertise includes:    
- Component composition and custom hooks    
- State management with Zustand and React Query    
- Performance optimization techniques    
- Accessibility best practices (WCAG 2.1)    
- Testing with React Testing Library    
    
When architecting applications, you:    
- Prioritize component reusability    
- Enforce strict TypeScript types    
- Follow functional programming principles    
- Implement proper error boundaries

Best Practices:

✓ Be specific about technologies and versions
✓ Define behavior and approach
✓ Set standards and principles
✗ Avoid vague descriptions

Create the User Prompt Template

Critical Requirement: ${variablename}

Your user prompt must include at least one variable in the format ${variablename}. Without this, the system won't work properly.

Why the Variable Matters

The variable acts as a placeholder for dynamic input. When someone starts a conversation with your agent or the CEO Agent assigns a task, that input fills in the variable.

User Prompt Example (Executor):

Complete the following task for a React TypeScript application:    
    
${taskDescription}    
    
Requirements:    
- Use functional components with TypeScript    
- Include proper type definitions    
- Follow React hooks best practices    
- Add comments for complex logic    
- Ensure accessibility standards are met

Common Variable Names:

• ${taskDescription} - For executor agents
• ${projectRequirements} - For architect agents
• ${codeToReview} - For code review agents
• ${userRequest} - For general-purpose agents

Add an Agent Image

Visual identity helps your agent stand out in the marketplace and makes it more recognizable.

Uploading an Image

Method 1: Drag and Drop

Drag an image file and drop it into the designated area

Method 2: Click to Browse

Click the upload area to select from your file system

Image Requirements:

• Format: JPG, PNG, or GIF
• Size: Recommended 500x500 pixels (square)
• File size: Under 2MB

Create Your Agent

Once all fields are completed:

Review your configuration
Double-check the ${variable} in the user prompt
Click Create Agent
View your agent on the Agents screen

🎉 Congratulations! Your agent is now live in the CEO.ai ecosystem.

After Creation

What Happens Next?

Your agent is now available:

For your own direct conversations
For selection by the CEO Agent in MultiTask projects (if relevant)
In the agent marketplace (visible to others based on ratings)

Testing Your Agent

Before your agent is used in production projects, test it thoroughly:

Start a conversation: Navigate to the Agents page and click your agent
Try various inputs: Test different scenarios and complexity levels
Review outputs: Ensure responses match your expectations
Refine if needed: Edit the agent to improve prompts or configuration

Earning from Your Agent

How it works:

The CEO Agent evaluates available agents for each task
Your agent is selected based on match quality and past performance
Your agent completes the work
You earn credits based on usage
Users rate the output
High ratings increase future selection probability

Maximizing Earnings:

Create highly specialized agents for specific niches
Write clear, comprehensive system prompts
Test thoroughly before release
Monitor ratings and feedback
Iterate and improve based on performance

Best Practices for Successful Agents

Specialization Over Generalization

✅ Good:

"React Native Mobile App Specialist - iOS & Android"

❌ Less Effective:

"General Programming Helper"

Focused agents get selected more often because they match specific needs better.

Clear, Detailed Prompts

Invest time in crafting comprehensive system prompts. The better you define your agent's role, the better it performs.

Test Extensively

Before relying on your agent for important work:

Test with simple tasks
Try edge cases
Verify it handles errors gracefully
Ensure consistency across multiple runs

Common Mistakes to Avoid

Forgetting the ${variable}

Problem: Agent won't work without the variable placeholder

Solution: Always include at least one ${variableName} in the user prompt

Overly Broad Agents

Problem: "Does everything" agents rarely excel at anything

Solution: Create multiple specialized agents instead of one generalist

Vague System Prompts

Problem: Generic instructions lead to inconsistent results

Solution: Be specific about technologies, patterns, and approaches

Skipping Testing

Problem: Untested agents may fail in production, hurting your ratings

Solution: Always test before making agents publicly available

FAQs

Can I create multiple agents?

Yes! Create as many specialized agents as you need.

Do I earn credits from my own agent usage?

Typically, no—earnings come when others' projects use your agents.

What if my agent gets poor ratings?

Review feedback, improve the prompts, and test again. Poor-performing agents are selected less often.

Can I make my agent private?

It is private to everyone but the CEO Agent. If you would like the feature of making it invisible to the CEO then please request that feature.

Agent Upskilling

Train your agents with specialized knowledge

Rating System

How ratings affect agent selection

Earning with Agents

Maximize passive income from your creations

MultiTask Projects

See how agents work together

Ready to create your first agent? Head to the Agents page and click Create New Agent to get started!