Kagenti Operator is a Kubernetes operator that automates the complete lifecycle management of AI agents, from building container images from source code to deploying and managing them in Kubernetes clusters.

The Kagenti Operator simplifies AI agent deployment by managing three Custom Resource Definitions (CRDs):

• Deploy from Image or Source — Use pre-built container images or build directly from GitHub repositories
• Automated Build Pipelines — Integrated Tekton pipelines with support for Dockerfile and Cloud Native Buildpacks
• Dynamic Agent Discovery — Kubernetes-native agent discovery through automatic indexing of agent metadata
• Flexible Configuration — Complete control over pod specifications, service ports, and environment variables
• Security Built-in — Support for private registries, secret management, and RBAC
• Multi-Framework Support — Works with LangGraph, CrewAI, AG2, and any A2A-compatible framework

graph TD; subgraph Kubernetes direction TB style Kubernetes fill:#f0f4ff,stroke:#8faad7,stroke-width:2px User[User/App] style User fill:#ffecb3,stroke:#ffa000 AgentCRD["Agent CR"] style AgentCRD fill:#e1f5fe,stroke:#039be5 AgentBuildCRD["AgentBuild CR"] style AgentBuildCRD fill:#e1f5fe,stroke:#039be5 User -->|Creates| AgentCRD User -->|Creates| AgentBuildCRD AgentController[Agent Controller] style AgentController fill:#ffe0b2,stroke:#fb8c00 AgentBuildController[AgentBuild Controller] style AgentBuildController fill:#ffe0b2,stroke:#fb8c00 Service_Service[Service] style Service_Service fill:#dcedc8,stroke:#689f38 Deployment_Deployment[Deployment] style Deployment_Deployment fill:#d1c4e9,stroke:#7e57c2 AgentPod[Agent Pod] style AgentPod fill:#c8e6c9,stroke:#66bb6a AgentCRD -->|Reconciles| AgentController AgentBuildCRD -->|Reconciles| AgentBuildController AgentController --> |Creates| Service_Service AgentController --> |Creates| Deployment_Deployment Deployment_Deployment --> |Deploys| AgentPod subgraph Tekton_Pipeline direction LR style Tekton_Pipeline fill:#e7f3e7,stroke:#73b473,stroke-width:1px Pull[1. Pull Task] style Pull fill:#e8eaf6,stroke:#5c6bc0 Build[2. Build Task] style Build fill:#fff3e0,stroke:#ffa726 Push[3. Push Image Task] style Push fill:#f3e5f5,stroke:#ab47bc Pull --> Build --> Push end AgentBuildController -->|Triggers| Tekton_Pipeline AgentBuildController -->|Saves Image URL on successful build| AgentBuildCRD AgentCRD -->|References| AgentBuildCRD end 

The operator separates build and deployment concerns:

• Agent CR manages deployment lifecycle and runtime configuration
• AgentBuild CR orchestrates the build process using Tekton pipelines

• Kubernetes cluster (v1.28+)
• kubectl configured to access your cluster
• Tekton Pipelines installed (for building from source)
• Container registry access (for building from source)

Using Helm:

# Install the operator using OCI chart helm install kagenti-operator \ oci://ghcr.io/kagenti/kagenti-operator/kagenti-operator-chart \ --version 0.2.0-alpha.19 \ --namespace kagenti-system \ --create-namespace

Option 1: From an existing container image

kubectl apply -f - <<EOFapiVersion: agent.kagenti.dev/v1alpha1kind: Agentmetadata: name: my-agent namespace: defaultspec: imageSource: image: "ghcr.io/kagenti/agent-examples/weather_service:v0.0.1-alpha.3" servicePorts: - port: 8000 targetPort: 8000 protocol: TCP name: http podTemplateSpec: spec: containers: - name: agent ports: - containerPort: 8000 env: - name: PORT value: "8000"EOF

Option 2: Build from source code

# First, create a build kubectl apply -f - <<EOFapiVersion: agent.kagenti.dev/v1alpha1kind: AgentBuildmetadata: name: my-agent-build namespace: defaultspec: mode: dev source: sourceRepository: "github.com/myorg/my-agent.git" sourceRevision: "main" sourceCredentials: name: github-token-secret buildOutput: image: "my-agent" imageTag: "v1.0.0" imageRegistry: "ghcr.io/myorg" imageRepoCredentials: name: ghcr-secretEOF# Then, deploy using the build kubectl apply -f - <<EOFapiVersion: agent.kagenti.dev/v1alpha1kind: Agentmetadata: name: my-agent namespace: defaultspec: imageSource: buildRef: name: my-agent-build servicePorts: - port: 8000 targetPort: 8000 protocol: TCP name: http podTemplateSpec: spec: containers: - name: agent ports: - containerPort: 8000EOF

# Check agent status kubectl get agents # Check agent build status kubectl get agentbuilds # View agent logs kubectl logs -l app.kubernetes.io/name=my-agent

See the config/samples directory for complete examples:

• weather-agent-image-deployment.yaml — Deploy from existing image
• weather-agent-build-and-deploy.yaml — Build and deploy from source
• helloworld-build-and-deploy-no-dockerfile.yaml — Use Cloud Native Buildpacks

We welcome contributions! See CONTRIBUTING.md for guidelines on:

• Reporting issues
• Submitting pull requests
• Development setup
• Testing requirements

Apache 2.0