kprompt.aiExperimental·The AI Runtime for Kubernetes
Tell Kubernetes what to do.
An AI-powered runtime that understands your cluster — operate on intent, investigate with evidence, and keep an always-on Observe agent watching namespaces without silent mutate. Plan before apply.
No API key · kind Observe demo · ~60 seconds
›
Intent in · context out · approve before apply
Observe agent
Always-on in the namespace. Same approval DNA.
The laptop CLI is reactive: you ask, it plans. The optional in-cluster Observe agent is continuous — watch, correlate, gated alert — without silent auto-heal. First AI Runtime surface inside Kubernetes.
- Laptop CLIReactive — you ask, it plans
- Observe agentAlways-on — watch → Incident → notify
Watch
Namespace-scoped watch on Pods, Events, and workloads — Role RBAC, not ClusterRole god-mode.
Correlate
Turns API noise into Incidents with evidence, health, and confidence — not a chat scroll.
Gate & alert
Slack / Discord / webhook only when severity and confidence clear the bar.
Propose, never silent mutate
Default Observe never applies. Optional Autopilot propose emits a PlanResult; apply stays human-gated.
› kprompt agent run -n payments --analyze --fetch-logs --health --heuristic
Observe: watching namespace payments
Incident: CrashLoopBackOff ×3 correlated
Evidence: OOMKilled · last restarts in 8m
Confidence: high · severity: warning
Alert gated — no mutate (Observe mode)
Laptop smoke: kprompt agent run -n payments --analyze --health --heuristic · kind walkthrough
Observe agent pipeline
Always-on, namespace-scoped. Batches by open Incident — not one LLM call per raw event. Mutate stays off.
- 01WatchPods · Events · …
- 02NormalizeTyped signals
- 03DedupeBurst collapse
- 04IncidentCorrelate window
- 05ContextLogs · metrics · GitOps
- 06AnalyzeLLM or --heuristic
- 07GateSeverity · confidence
- 08NotifySlack / webhook
Runtime story
Understand. Investigate. Change. Improve.
An intelligent runtime between your intent and the cluster — not a chatbot that emits kubectl, and not a monitoring dashboard. Pair the reactive CLI with an optional Observe agent when you want always-on watch in a namespace.
- Human intent
- kprompt runtime
- Kubernetes
- Workloads · deps · signals
01
Intent
State the outcome — scale a workload, ask why something is failing, or analyze the cluster. You are not writing kubectl.
02
Understand
kprompt reads Kubernetes state, relationships, events, and available signals (logs, metrics, traces) in context.
03
Reason
It decides what matters, what is related, what looks abnormal, and what is worth investigating or improving.
04
Act / recommend
Execute operations, investigate problems, explain findings, or surface risks — with a reviewable plan before apply.
What you can do
One runtime. Many outcomes.
Investigation is a capability — not the whole product. Operate with the CLI, keep an Observe agent on the namespace, analyze with the same approval DNA.
Operate
Shippedkprompt "scale checkout-api to 10"
Intent → plan → approve → apply
Observe
Shippedkprompt agent run -n payments --analyze --health
Always-on agent — watch, correlate, gated alert; no silent mutate
Investigate
Shippedkprompt why checkout-api
Evidence-backed findings, not chat guesses
Analyze
Shippedkprompt "optimize my cluster"
Inventory, idle, rightsizing, HPA hints
Differentiation
Context, not commands or charts alone.
kprompt is a runtime that reasons about your Kubernetes environment — not a chat REPL, not a thin kubectl wrapper, and not an AIOps dashboard.
Generic AI chatbot
They might
“How do I create a PDB?”
kprompt
Questions about your cluster — which workloads look risky, idle, or misconfigured — using live context, not docs alone.
Kubernetes CLI wrapper
They might
“restart checkout” → a kubectl command
kprompt
Understand environment state first, then propose a reviewable PlanResult before anything applies.
In-cluster agent platform (e.g. kagent)
They might
Agents as CRDs · MCP · A2A · GitOps agent product
kprompt
PlanResult ops CLI + optional Observe notify — not a multi-agent control plane. Keep kagent when agents are the product.
AI / MCP gateway (e.g. agentgateway)
They might
LLM · MCP · A2A traffic policy on Gateway API
kprompt
Ops compiler at the keyboard — not a data-plane proxy. Keep agentgateway when every agent hop needs a governed edge.
Traditional observability
They might
“CPU: 94%”
kprompt
Connect symptoms with related workloads, dependencies, and signals when they are available — not a single chart.
For operators who already know Kubernetes
You already know Kubernetes. kprompt helps you see more of it.
Experts know kubectl, PDBs, HPA, topology spread, and common failure modes. What still hurts at scale is gathering and correlating context across every workload, dependency, event, metric, and risk — continuously.
kprompt augments that expertise. It does not pretend to replace it.
Philosophy
Kubernetes is already powerful.
The problem is the amount of context humans need to understand and operate it. kprompt exists to provide that context — so you can work at the level of intent: operate, investigate, and improve.
Not primarily because Kubernetes commands are hard. Because clusters are large, related, and noisy — and your time is finite.
CLI interface
One binary into the runtime.
Natural language in. Context-aware plan out. Mutate only after you approve — operate, investigate, and analyze from the same CLI.
› kprompt "scale api to 10" --approve --wait
✓ Scaling deployment/api 3 → 10
Waiting for rollout…
✓ Ready
Signals & stack
Connect the dots — not fifty connectors
Kubernetes plus the logs, metrics, and traces you already run enrich the runtime. Orchestrated, not replaced. No bespoke Redis/MySQL product connectors.
Providers
Your LLM, your key
Local Ollama ($0) or BYOK — kprompt does not sell API keys.
kprompt orchestrates Kubernetes, Helm, Argo, Prometheus, OpenTelemetry, Grafana, Tekton, KEDA, Istio, Crossplane, Flux. Natural-language plans use Ollama, OpenAI, Anthropic, Gemini, Groq, xAI, Mistral, DeepSeek, Moonshot, OpenRouter, Together, or any OpenAI-compatible gateway. It does not replace these tools or sell inference.
Questions operators ask first
Straight answers about approval, credentials, and what is still experimental.
What is an AI Runtime for Kubernetes?
kprompt is an AI-powered runtime that understands your Kubernetes environment and turns intent into context-aware operations, investigations, and recommendations. You describe the outcome; it gathers relevant cluster context, reasons about what matters, and produces a reviewable PlanResult before any mutation. It is not a ChatGPT wrapper, a kubectl command generator, a monitoring-only AIOps dashboard, or a silent auto-healer.
ArchitectureIs kprompt only for incidents and AI SRE?
No. Investigation (investigate / why / timeline / impact) and the optional Observe agent are one capability lane. The same runtime also covers day-2 operate (deploy, scale, rollback), explain/read, optimize and audit reports, dependency graphs, and multi-tool routes — always under plan → approve for mutations.
OverviewWhy would an experienced Kubernetes engineer use it?
You already know kubectl, PDBs, HPA, and failure modes. What still hurts at scale is gathering and correlating context across workloads, dependencies, events, metrics, and risks. kprompt augments that expertise by reducing manual correlation — it does not pretend to replace it.
Beyond AI kubectlHow is kprompt different from a chatbot, CLI wrapper, or dashboard?
A chatbot answers how-to docs; kprompt reasons about your live cluster. A CLI wrapper translates English into kubectl; kprompt understands environment state before proposing a PlanResult. A dashboard shows isolated signals; kprompt connects related evidence when signals are available. Same word “AI,” different contract.
Kubernetes AI tools comparedDo I need Prometheus, Loki, or OpenTelemetry?
No for core Kubernetes operate and many reads. When Prom, OTel, Grafana, or similar backends are configured, they enrich understanding (perf explain, optional graph edges, dashboards). Missing backends degrade honestly — kprompt does not invent metrics or claim per-database product connectors.
IntegrationsWhat is the Observe agent?
An optional in-cluster Observe agent that watches a namespace, correlates Pods/Events into Incidents, and sends gated Slack/Discord/webhook alerts. Default mode never mutates. Optional Autopilot propose emits a PlanResult; apply stays human-gated. The laptop CLI still works without any agent — try kprompt demo for a $0 kind walkthrough.
Observe agent docsAI Runtime vs AI Gateway vs Agent Platform — which is kprompt?
kprompt is an AI Runtime for Kubernetes ops: PlanResult → safety → approve. An AI Gateway (e.g. agentgateway) governs LLM/MCP/A2A traffic. An Agent Platform (e.g. kagent) runs Agents as CRDs. Same vocabulary, three layers — see the hub post for the triangle and deep links.
Runtime vs Gateway vs PlatformHow is kprompt different from agentgateway?
agentgateway is an AI-native proxy for LLM, MCP, and A2A traffic (Gateway API). kprompt is a plan-before-apply ops CLI (and optional Observe notify) — natural language → PlanResult → approve. kprompt mcp serve is IDE interop, not a cluster MCP gateway. Keep agentgateway for traffic governance; choose kprompt for gated day-2. Many teams use both.
kprompt vs agentgatewayIs kprompt an agentgateway alternative?
Yes for the job “gated Kubernetes day-2 / SRE without standing up an AI/MCP gateway.” No for the job “LLM/MCP/A2A Gateway API data plane.” See the agentgateway alternatives hub for a job-based map.
agentgateway alternativesHow is kprompt different from kagent?
kagent (CNCF Sandbox) is a Kubernetes-native agent platform: Agents as CRDs, MCP tools, A2A, GitOps. kprompt is a plan-before-apply ops CLI (and optional Observe notify agent) — natural language → PlanResult → approve. Overlapping SRE demos, different products. Keep kagent when you need an agent control plane; choose kprompt when you need a refuse-able day-2 plan. Many teams use both.
kprompt vs kagentIs kprompt a kagent alternative?
Yes for the job “gated Kubernetes day-2 / SRE without standing up Agents-as-CRDs.” No for the job “CNCF agent platform with MCP/A2A.” See the kagent alternatives hub for a job-based map.
kagent alternativesDoes kprompt apply changes to my cluster automatically?
No. Every mutating intent compiles into a PlanResult that lists the resources, diffs, and risk level before anything runs. On a TTY you confirm with y/N; in scripts you pass --approve explicitly. Wipe-class prompts are hard-denied instead of planned, and Autopilot is propose-only by default (ADR-0015) — it never applies silently.
Safety modelShould I replace kubectl or K9s with kprompt?
No. kubectl remains the precise API client and scripting language; K9s remains the best live terminal UI for watching rollouts and hopping between Pods. kprompt sits beside them when the bottleneck is correlating intent and cluster context — not exact flags or a resource browser.
kubectl vs K9sWhere do my kubeconfig and LLM keys live?
On your machine. kprompt reads your existing kubeconfig locally. For NL plans use local Ollama ($0, no key) or your own cloud provider key (BYOK). Cluster credentials are not uploaded to a control plane by default.
Providers and keysCan I use kprompt in CI/CD pipelines?
Yes, for plan generation and gating. kprompt emits a stable JSON PlanResult with intent, resources, diffs, and risk, so a pipeline can fail on high-risk plans or post the plan to a pull request before a human approves apply.
CI / PlanResult JSONIs kprompt production ready?
It is labelled experimental on purpose. Plans can be wrong or incomplete outside the hard-deny rules, so read every plan before apply, start on kind or a sandbox cluster, and treat --approve with care.
Roadmap and honest limitsIs kprompt free and open source?
Yes. The CLI is Apache-2.0 licensed on GitHub and stays free. There is an optional Team surface for organisations that already have access, but there is nothing to buy as a public signup product today.
Install the CLIRun the runtime where your cluster is
Install the CLI, try the $0 Observe walkthrough on kind, then operate with intent or Helm the agent into a namespace. Prefer non-production first — experimental software.
- 01Install the CLI
- 02Clone kprompt-examples
- 03make walkthrough ($0 Observe)
- 04Ollama or BYOK for NL (optional)
Apache-2.0 · open source · kind Observe demo