​

Agent Servers Overview

​

What Are Agent Servers?

• 🧠 Generate complex workflows from natural language descriptions
• ⚙️ Execute multi-step processes with error handling and retries
• 🔄 Coordinate multiple agents for complex problem-solving
• 📊 Monitor and optimize workflow performance
• 🛡️ Provide enterprise features like authentication and auditing

​

Available Agent Servers

​

Agent Servers vs MCP Providers

​

Architecture Comparison

​

Getting Started with Agent Servers

​

Quick Decision Guide

​

Minimum Requirements

• Python 3.10+ with virtual environment
• API Keys: Kubiya + AI provider (Together AI, OpenAI, etc.)
• Memory: 2GB+ for local development
• Network: Outbound HTTPS access for AI APIs

​

Google ADK (Recommended)

​

Key Features

​

Quick ADK Setup

# Install dependencies
pip install kubiya-workflow-sdk google-adk

# Set environment variables
export KUBIYA_API_KEY="your-key"
export TOGETHER_API_KEY="your-key"

# Start ADK server
python3 workflow_sdk/adk_orchestration_server.py

​

Custom Agent Servers

• Custom AI models or providers
• Specialized workflow types for your domain
• Unique security requirements
• Custom integrations with proprietary systems

​

Basic Custom Server

from kubiya_workflow_sdk.server import OrchestrationServer
from kubiya_workflow_sdk.providers import CustomProvider

class MyCustomProvider(CustomProvider):
    async def compose(self, task: str, mode: str):
        # Your custom workflow generation logic
        workflow = await self.generate_workflow(task)
        
        if mode == "act":
            result = await self.execute_workflow(workflow)
            return result
        
        return workflow

# Create and start server
server = OrchestrationServer(provider=MyCustomProvider())
server.start(port=8001)

​

Integration Examples

​

Frontend Integration

// Detect agent server capabilities
const serverResponse = await fetch('http://localhost:8001/discover');
const capabilities = await serverResponse.json();

if (capabilities.server.capabilities.orchestration) {
  // This is an agent server - can handle complex workflows
  const workflowResponse = await fetch('http://localhost:8001/compose', {
    method: 'POST',
    body: JSON.stringify({
      prompt: "Deploy my application to staging",
      mode: "act"
    })
  });
}

​

MCP Integration

# Expose agent server capabilities through MCP
@mcp.tool()
def execute_complex_workflow(description: str, mode: str = "plan"):
    """Execute a complex multi-step workflow"""
    return agent_server.compose(task=description, mode=mode)

​

Production Considerations

​

Security

• API Authentication: Always use API keys in production
• Rate Limiting: Implement request quotas
• Input Validation: Sanitize all user inputs
• Audit Logging: Track all workflow executions

​

Scalability

• Load Balancing: Multiple server instances
• Caching: Workflow templates and responses
• Monitoring: Health checks and metrics
• Resource Limits: Memory and CPU constraints

​

Reliability

• Error Recovery: Graceful failure handling
• Retry Logic: Automatic error recovery
• Health Monitoring: Continuous server monitoring
• Backup Strategies: Data persistence and recovery

​

Next Steps