Chat with AI

The AI chat lets you interact with the mogenius AI directly through a conversational interface. You can ask questions, investigate issues, and give instructions — the assistant has access to your cluster and uses its tools to find answers and take action, reasoning step by step in a think → act → observe loop.

​

Opening the chat

Open the chat from the chat button in the platform header — it appears in a sidebar next to your current view. You can also open it as a separate window to keep the conversation visible while you work elsewhere. The chat is available at both the Cluster and Workspace level, and automatically scopes its responses to the context you’re working in.

​

What you can do

The assistant understands natural language and can use its tools to interact with your cluster on your behalf. Common uses include:

• Investigating issues such as misconfigured resources, unreachable applications, or networking problems.
• Querying cluster resources like Pods, Deployments, Services, and ConfigMaps.
• Getting guidance on resource optimization, node allocation, and resource limits.
• Reviewing and drafting Kubernetes manifests or Helm values.
• Troubleshooting interactively by feeding the assistant additional context.

​

Chat modes

A mode selector tailors how the assistant responds:

• Ask — Answer questions and explain, without proposing changes.
• Code — Focus on writing and editing manifests, values, and configuration.
• Plan — Break a task down into a step‑by‑step plan before acting.

You can clear the conversation at any time to start fresh.

​

Choosing a model

A model picker lets you switch between the models marked as chat enabled in AI Models & Setup. The cluster default is preselected. Switching models starts a fresh connection for the next message.

​

Tools and reasoning

The chat uses the same tools as agents. As the assistant works, it shows which tools it is calling so you can follow its reasoning:

• Kubernetes — Query resources, read logs, describe Pods, inspect events, and retrieve manifests in real time.
• Helm — List releases, inspect chart values, and review release history.
• Custom tools — Any custom MCP servers and integrations connected to your cluster.

See Tools & Custom MCP Servers for the full picture.

​

Streaming and cancelling

Responses stream in as they are generated, with an indicator while the assistant is thinking. If a response is heading in the wrong direction, use Cancel to stop it immediately and refine your prompt. Each message offers quick actions — copy, retry, and edit & resend — plus feedback controls, and code blocks have their own copy button.

​

Permissions

The assistant operates under the same permissions as the logged‑in user. If your role grants View access to specific namespaces, the chat can only query resources in those namespaces. It never exposes data beyond what you are authorized to see.

​

Token usage

Chat conversations count toward the token limits of the model in use. A usage indicator shows how much of the budget has been consumed. Once a limit is reached, further messages are queued until it resets. See token and budget limits.

​

Session persistence

Chat sessions are preserved across page reloads and browser refreshes, scoped to your user and the current organization, cluster, and workspace. You can close the chat, navigate elsewhere, and return to an ongoing conversation without losing context.