Skip to content
GovOps

Foundations

Use Case: Agentic Workloads

1 min readEdit on GitHub

#Scenario: Natural Language Interface to Cloud-Native Infrastructure

This use case describes a scenario where a human operator uses a natural language interface (NLI) to deploy an application to a cloud-native environment (e.g., Kubernetes). This scenario involves multiple trust domains and is a good example of the type of complex, agentic workload that GovOps is designed to govern.

#Actors

  • Human Operator: The user interacting with the NLI.
  • Natural Language Interface (NLI) Agent: An agent that interprets the operator's commands.
  • Model Context Protocol (MCP) Server: A server that provides the NLI agent with the necessary skills and context.
  • Kubernetes Cluster: The target environment for the application deployment.
  • Identity Provider (IDP): The provider that issues tokens to the various actors.

#Flow

  1. The Human Operator issues a command to the NLI Agent, e.g., "Deploy the latest version of the webapp application."
  2. The NLI Agent calls the MCP Server to get the necessary skills to perform this operation.
  3. The MCP Server returns the relevant skills, which may involve interacting with the Kubernetes Cluster.
  4. The NLI Agent, acting on behalf of the operator, makes a request to the Kubernetes Cluster to deploy the application.
  5. The Kubernetes Cluster and its components (e.g., API server, custom admission controllers) evaluate the request against their policies.

#GovOps in Action

In this scenario, GovOps can be used to govern the interactions between the different components:

  • Capability Definition: Each operation (e.g., deploy-application, get-mcp-skill) is defined as a capability in the Authorization Capability Catalog (ACC).
  • Policy Enforcement: Policies are enforced at multiple points:
    • The MCP Server may have policies that restrict which agents can access which skills.
    • The Kubernetes Cluster will have policies that restrict who can deploy applications.
  • Runtime Observability: As each policy is evaluated, an observable event is generated. This provides a complete, real-time audit trail of the entire operation, from the initial NLI command to the final deployment.
  • Metrics: The observable events can be used to generate metrics, such as the number of successful deployments, the number of denied requests, and the denial ratio trend.