Skip to main content
Version: 1.37

AI Agent Fleets - Getting Started

Welcome to AI Agent Fleets! This guide will help you get up and running quickly with AI-powered development agents in your Okteto environments.

📋 Prerequisites​

Before you begin:

  • ✅ Your administrator has enabled AI Agent Fleets for your organization
  • ✅ You have access to the Okteto Dashboard
  • ✅ You can see the Agents tab in your dashboard

Don't see the Agents tab? Contact your administrator to request access.

🚀 Your First Agent in 3 Minutes​

Step 1: Open the Agents Tab​

Navigate to the Agents tab in your Okteto Dashboard.

Step 2: Choose Your Starting Point​

You have two options:

Option A: Work with an Existing Repository​

  • Paste your repository URL (e.g., https://github.com/yourorg/yourrepo)
  • The agent will clone and analyze your codebase
  • Perfect for adding features or fixing bugs

Option B: Start Fresh​

  • Select "New Project"
  • Describe what you want to build
  • The agent will create everything from scratch

Step 3: Describe Your Task​

Write a clear, specific prompt. For example:

Create a Python FastAPI application with:
- User registration endpoint with email validation
- Login endpoint returning JWT tokens
- PostgreSQL database with SQLAlchemy
- Automated tests for all endpoints

Step 4: Launch and Monitor​

  1. Press Enter or click Launch Agent
  2. Watch the real-time logs as your agent:
    • Sets up the environment
    • Installs dependencies
    • Writes code
    • Runs tests
    • Validates the implementation
  3. Click on Open Editor to view an embedded Visual Studio Code where you can monitor and audit all code changes

Step 5: Review and Deploy​

Once complete:

  • Preview: Click any generated endpoints to see your app live
  • Review: Check the code changes and test results
  • Accept: Just ask the agent to create a pull request

💡 Quick Tips for Success​

Writing Effective Prompts​

Do This ✅

"Add a REST endpoint at /api/users/:id/avatar that accepts 
image uploads (JPEG/PNG only, max 5MB), resizes to 200x200, 
stores in S3, and returns the CDN URL"

Not This ❌

"Add image upload"

Common First Tasks​

Start with these proven scenarios:

  1. Add an API Endpoint

    Add a health check endpoint at /health that returns 
    server status, database connectivity, and uptime

  2. Create a New Service

    Create a Node.js microservice for sending emails with 
    SendGrid, including templates and retry logic

  3. Fix a Bug

    Debug why the user authentication is failing with 
    401 errors after 15 minutes and implement a fix

  4. Add Tests

    Write comprehensive unit tests for the OrderService 
    class with at least 80% coverage

📖 Download the Complete Prompt Guide​

Want to master AI Agent Fleets? Download our comprehensive prompt guide with:

  • The 5-pillar framework that transforms frustrating AI interactions into production-ready solutions
  • Advanced techniques for complex tasks

📥 Download the AI Agent Fleets Prompt Guide (PDF)

🎯 Best Practices​

1. Start Small, Think Big​

  • Begin with simple, well-defined tasks
  • Build confidence with successful completions
  • Gradually increase complexity

2. Provide Context​

Good prompts include:

  • What: The specific feature or fix needed
  • Where: Relevant files, endpoints, or services
  • How: Technical requirements or constraints
  • Why: Business logic or user needs (when relevant)

3. Leverage Parallel Agents​

Run multiple agents simultaneously:

  • Agent 1: Adding new features to the frontend
  • Agent 2: Optimizing database queries
  • Agent 3: Writing documentation

4. Review Before Merging​

Always:

  • Run the preview environment
  • Check test results
  • Review code for security and best practices
  • Validate against your requirements

5. Iterate and Refine​

If the first result isn't perfect:

  • Provide specific feedback
  • Ask for modifications
  • Guide the agent with additional context

🔧 Common Workflows​

Feature Development Workflow​

  1. Describe the feature with acceptance criteria
  2. Let agent implement the initial version
  3. Test in preview environment
  4. Request adjustments if needed
  5. Create PR when satisfied

Bug Fixing Workflow​

  1. Describe symptoms and error messages
  2. Let agent investigate and propose fix
  3. Verify fix in preview environment
  4. Run regression tests
  5. Deploy fix via PR

Refactoring Workflow​

  1. Specify refactoring goals (performance, readability, etc.)
  2. Define constraints (maintain API compatibility, etc.)
  3. Review changes carefully
  4. Validate all tests still pass
  5. Deploy incrementally

🚨 When to Use AI Agent Fleets​

Perfect For ✅​

  • CRUD operations
  • API endpoints
  • Data transformations
  • Test writing
  • Bug fixes with clear symptoms
  • Documentation updates
  • Boilerplate code
  • Database migrations

Think Twice About 🤔​

  • Core authentication/authorization logic
  • Payment processing
  • Complex architectural decisions
  • Cryptographic implementations
  • Performance-critical algorithms

Not Recommended For ❌​

  • Production secrets/credentials
  • Regulatory compliance code (without review)
  • Mission-critical security features (without expert review)

🔍 Understanding Agent Capabilities​

AI Agents can:

  • ✅ Read and understand your entire codebase
  • ✅ Install packages and dependencies
  • ✅ Create new files and modify existing ones
  • ✅ Run commands and scripts
  • ✅ Execute and debug tests
  • ✅ Access Okteto environment variables and configs
  • ✅ Interact with databases
  • ✅ Call external APIs
  • ✅ Generate comprehensive documentation

AI Agents cannot:

  • ❌ Access production environments
  • ❌ Bypass your CI/CD pipeline
  • ❌ Merge code without approval
  • ❌ Access secrets not provided to the environment
  • ❌ Modify Okteto platform settings

🆘 Quick Troubleshooting​

Agent is taking too long​

  • Break complex tasks into smaller steps
  • Provide more specific instructions
  • Check to see if your environment is running low on resources

Agent doesn't understand my codebase​

  • Ensure repository is accessible
  • Provide context about architecture
  • Reference specific files/patterns

Tests are failing​

  • Specify test requirements clearly
  • Provide example test cases
  • Check environment dependencies

Can't see preview environment​

  • Wait for deployment to complete
  • Check service health endpoints
  • Verify port configurations

💬 Get Help​

🚀 What's Next?​

Now that you're familiar with the basics:

  1. Download the Prompt Guide for advanced techniques
  2. Try the examples in your test environment
  3. Join the community to share tips and learn from others
  4. Experiment with parallel agents for complex projects

Ready to accelerate your development? Launch your first agent now!

Pro Tip: Keep this guide bookmarked and refer to the prompt guide for specific use cases. The more specific your prompts, the better your results!