monkMonk or Kubernetes?
Kubernetes already reconciles desired state and recovers workloads. The decision is how much of the surrounding application platform you want your team to build and operate.
Kubernetes fits when…
You need the Kubernetes ecosystem, custom controllers and Kubernetes-native tooling, or your team already runs an effective platform on it.
Monk fits when…
You want an agent-driven workflow spanning application deployment, supported cloud services, connections and operations without first assembling a Kubernetes platform.
Try it alongside your current setupWhat changes for your team
| Responsibility | Kubernetes | Monk |
|---|---|---|
| Deploy | Kubernetes schedules and runs container workloads. Builds, ingress, external services and delivery workflows come from the platform you configure. | Plans the app, infrastructure and service connections. Deploys after the required approvals. |
| State | The API and controllers hold and reconcile desired state. Operators and tools such as Crossplane can extend this to external resources. | The orchestrator retains the application and resource model across agent sessions. |
| Recovery | Controllers restart and replace workloads under configured policies. Self-healing is a Kubernetes capability. | Configured workload recovery is automatic. Watcher diagnoses persistent issues and proposes fixes for approval. |
| Controls | RBAC, admission policies and delivery tooling can enforce team rules independently of an AI agent. | The orchestrator enforces permissions and validation. Deployment plans have a separate approval screen. |
| Existing apps | Move or deploy workloads onto the cluster and integrate them with your platform’s services and policies. | Start with a separate deployment from source. Plan data transfer and traffic cutover explicitly. |
| Costs | Account for cluster infrastructure, managed control-plane fees where applicable and the tools and people operating the platform. | Your cloud bills you directly. Monk and your coding agent are separate costs. |
Sources and comparison scope
Comparison baseline reviewed 12 September 2026. Workflow comparison, not a performance benchmark. Capabilities depend on configuration and plan. Primary documentation is linked below.
Give both the same job.
Deploy a frontend, API, database, export worker and storage. Create a task, run an export and retrieve the file.
With Kubernetes
Package the task app for your Kubernetes platform. Provide a database and storage, configure identities and connections, and deploy web, API and job workloads. Your controllers handle configured recovery. Verify the same export workflow.
With Monk
Ask Monk to plan the app and its dependencies. Review resources, credentials and costs, then approve deployment. Create a task and retrieve its exported CSV. For an existing app, start in a separate environment.
Before you switch.
Monk does not replace the Kubernetes ecosystem one feature at a time. Compare your actual platform requirements and supported integrations before moving an existing application.
Can I evaluate Monk without replacing production?
Monk builds fresh infrastructure from your codebase. Your existing deployment stays untouched. Test the new environment before you switch. Allow for duplicate cloud costs during evaluation. Data transfer and cutover are separate decisions.
Will Monk take over my current infrastructure?
The default path is a new deployment from source. Do not have two infrastructure controllers manage the same resource. Keep ownership boundaries explicit during evaluation.
Try it with your application.
Start from your repository. Review the resources and costs before you deploy.