Skip to content
Kubeterm Logo Kubeterm
Download

AI Assistance

Kubeterm includes built-in AI chat designed around real Kubernetes workflows. The AI can read your cluster state directly and surface results as interactive views inside the conversation.

Supported AI Providers

Bring your own API key for any of the supported providers. The AI settings let you configure multiple providers and switch between them:

  • OpenAI (GPT models via /v1/chat/completions)
  • Anthropic (Claude, including the latest Opus models)
  • Google Gemini
  • DeepSeek

Each provider can be configured with a custom base URL, model name, temperature, and max token limit. Provider credentials are stored in secure storage. On Apple devices, AI provider settings can be synced via iCloud so you don't need to re-enter them on each device.

Chat Capabilities

  • Cluster-Aware Context: Select which cluster to work in from the chat panel. When you open AI chat from inside a resource view (logs, resource list, etc.), the cluster and resource context are passed automatically so you don't need to re-specify what you're looking at.
  • Embedded Results: AI responses can include live resource lists, describe output, YAML, and logs directly inside the conversation — no need to switch screens.
  • Tool Use / kubectl Operations: The AI can perform read operations against the cluster API (listing resources, describing objects, fetching logs) to answer your questions with actual cluster data. Tool calls and their results are rendered inline in the order they happen, so you can follow exactly what the assistant did.
  • Web Search: The assistant can search the web mid-conversation to pull in current documentation, changelogs, and error explanations alongside what it reads from your cluster.
  • Cluster Events & Version: Ask what recently happened in a namespace, or which Kubernetes version a cluster runs, and get the answer from the live cluster rather than a guess.
  • File Transfers: Ask the assistant to copy a file out of a pod for inspection, or push one in — the same transfer as kubectl cp, driven from the conversation.
  • Explain & Health Check: Every resource page has an action that hands the resource to the assistant for an explanation or a health assessment, without composing a prompt yourself.
  • Tool Policy: You control which kubectl operations the AI is allowed to perform. Restrict it to read-only or allow specific write actions.
  • Streaming Responses: Responses stream in token by token. Models that support reasoning (thinking) show the reasoning steps before the final answer.
  • Context Compression: Long conversations are automatically compressed to stay within model context limits while preserving the key details.
  • Persistent History: Chat history is preserved across sessions for each cluster context.
  • Failure Analysis: Analyze a failing resource directly from the resource detail view — the AI receives the resource manifest and recent events to help diagnose the issue.