How It Works

Kaigents is built on a Kubernetes-native architecture that treats AI agents as first-class resources. If you know how to operate a Kubernetes cluster, you already know how to operate Kaigents.

CRD Model

All core components of Kaigents are implemented as Kubernetes Custom Resource Definitions (CRDs):

  • Agent - Defines the AI agent's behavior and configuration
  • Skill - Process-adherence playbook that constrains agent output
  • Task - Individual work units that agents execute
  • Process - Workflow definition for multi-step agent operations
  • Run - Execution instance of a process

This declarative approach enables GitOps workflows and makes it easy to version control your AI agent configurations.

Execution Flow

The execution process follows a predictable flow:

  1. Define an Agent with Skills and configuration
  2. Create a Process that orchestrates Tasks
  3. Submit a Run to execute the Process
  4. Agent executes Tasks with durable, long-running capabilities
  5. Results are stored in S3-compatible artifact storage

This flow ensures that AI agent workloads are observable, governable, and portable across environments.

Benefits of Kubernetes-Native Approach

  • Declarative definitions that are version-controlled
  • GitOps-friendly workflows
  • Integration with existing Kubernetes tooling
  • Scalability and resource management through Kubernetes
  • Observability through Kubernetes monitoring tools

This architecture makes Kaigents a natural fit for organizations already invested in Kubernetes infrastructure.