Skip to main content
Chat is RubixKube’s conversational surface. No kubectl memorisation, no grepping logs, no clicking through five dashboards. Ask the question you would ask a senior engineer, get a cited answer backed by the Agent Mesh. This tutorial set takes you from first query to daily driver. Start at Basics if you are new, skip ahead if you are not.

​
Why Chat

Natural language

Plain English. No query language to learn.

Answers with evidence

Every claim cites the underlying signal. Verify in one click.

Context aware

The session remembers the thread. Follow-ups do not start from scratch.

Multi-agent backend

Observer, RCA Pipeline, Memory, and Guardian all feed the answer.

Works across environments

Same prompts work for Kubernetes, AWS, GCP, and VMs in one workspace.

Runs skills on demand

Invoke any enabled skill with a slash command or by intent.

​
The difference, in two lines

The old way on Kubernetes: 
kubectl get pods -n production
kubectl describe pod api-gateway-xyz
kubectl logs api-gateway-xyz
kubectl top pod api-gateway-xyz
# five more commands, then stitch it together in your head
The Chat way: 
Why is api-gateway slow in production?
One cited answer, no context-switching. The same prompt works whether api-gateway is a Kubernetes pod, an ECS task, a GCE instance, or a process on a VM.

​
The tutorial series

Chat Basics

Open the interface, type your first query, read the response. About ten minutes.

Troubleshooting with Chat

Real debug flows: OOMKilled, ImagePullBackOff, CrashLoopBackOff, latency spikes. About fifteen minutes.

Advanced Chat

Role-shaped workflows for SREs, DevOps, and platform engineers. About twenty minutes.

Chat for cost analysis

Rightsizing, anomaly hunting, weekly budget reviews, all through Chat. About fifteen minutes.

​
The four prompt shapes that work

Use for the daily read.
What is the health of every production environment?
Which workloads are degraded right now?
Any hosts above 80% memory in the last hour?
Use when something just broke.
What changed in the checkout service in the last hour?
Which deployments rolled back in the last 24 hours?
Show me configuration drifts on prod-cluster-us-east this week.
Use to pull a specific slice for a deep dive.
Why is the payments API slow in ap-south-1?
Which pods restarted most in the last 24 hours, grouped by namespace?
Top five CPU consumers on host-42 in the last hour.
Use for weekly rightsizing.
Which services have the most headroom on CPU limits?
RDS instances with the lowest average utilisation this month.
GCE instances over-provisioned on memory, sorted by spend.

​
Slash commands worth knowing

The slash menu shows up when you type / in the chat box.
CommandPurpose
/environmentsSwitch which environment the next question runs against
/newStart a fresh session when the context gets muddy
/resumePick up an earlier investigation where you left off
/renameName the session so you can find it later
/sendPush the last shell output into chat
/skillsInvoke a saved skill (runbook) against the current question

​
Where to start

Work through the series in order:
  1. Chat Basics, ten minutes.
  2. Troubleshooting with Chat, fifteen minutes.
  3. Advanced Chat, twenty minutes.
  4. Chat for cost analysis, fifteen minutes.
About an hour, end to end, with one real cluster.

​
Quick reference

Five prompts that cover most of daily use:
  1. What is failing? - morning triage.
  2. Why is <service> slow? - focused investigation.
  3. What changed in <namespace> in the last hour? - post-deploy check.
  4. Validate this (with YAML uploaded) - pre-deploy safety.
  5. Explain <concept> - learn a Kubernetes or AWS concept using your own resources as examples.
Fastest way to open Chat: ⌘K in the console, or rubix in your terminal.

​
Common questions

Yes. The same prompts work for Kubernetes, AWS, GCP, and Linux VMs. Use /environments to scope to one, or ask broadly and Chat will include every environment in the workspace.
The SRI Agent that powers Chat runs in RubixKube Cloud. The Observer in your environment provides signals. Raw logs stay with you unless you explicitly enable broader streaming.
Chat can recommend and draft changes. Applying anything requires explicit approval and is bounded by Guardian policies. See How to Automate Incident Remediation.
Yes. Install with npm install -g @rubixkube/rubix, launch with rubix. Same agent, same skills, terminal-native UX.

How to add custom agent skills

Turn your best prompts into reusable skills.

How to Monitor Infrastructure Health

Dashboard-first view of the same data.