# KubeIntellect

> KubeIntellect is an AI-powered Kubernetes management assistant. It diagnoses cluster failures, performs root cause analysis, automates cluster operations, and answers questions about your Kubernetes workloads through natural language — with human-in-the-loop safety gates.

KubeIntellect is built as a multi-agent LLM system that connects to your Kubernetes cluster via kubeconfig, runs kubectl commands, and synthesizes findings into actionable diagnoses with suggested fixes. It is free for individuals to use. KubeIntellect is also open source, licensed AGPL-3.0-or-later with a commercial license available. The CLI tool is called `kq` and is installable with `pip install kube-q`; the server is `pip install kubeintellect` or the container image `ghcr.io/mskazemi/kubeintellect`.

## Product

- [Homepage](https://kubeintellect.com): Product overview, features, and free signup
- [Architecture](https://kubeintellect.com/architecture): How KubeIntellect works — a LangGraph supervisor orchestrating 13 specialized agents and a code-generator for dynamic tool synthesis, reasoning across live cluster evidence, with a human-in-the-loop approval gate before any write/delete/exec on the Kubernetes API
- [Demo](https://kubeintellect.com/demo): Live interactive demo — no sign-up required, works against a shared demo cluster
- [Documentation](https://kubeintellect.com/docs): Full product documentation including CLI reference
- [Research](https://kubeintellect.com/research): The peer-reviewed paper, evaluation results, open-source repositories, how to cite KubeIntellect, and a build-in-public changelog. KubeIntellect is peer-reviewed and published in the Journal of Grid Computing (Springer), vol. 24, no. 3, 2026 — DOI 10.1007/s10723-026-09837-6, by Mohsen Seyedkazemi Ardebili and Andrea Bartolini. Peer-reviewed results: 93% query resolution (186/200), +25 points over a GPT-4o baseline, 81.8% runtime tool-synthesis success, 7–10s per query, on a live four-node cluster (170 pods, 18 namespaces).

## Open source & contributing

KubeIntellect is open source under AGPL-3.0-or-later (dual-licensed; a commercial license is available for closed or SaaS use). Development happens in public on GitHub at github.com/MSKazemi/kubeintellect, and the project is actively seeking contributors. Contributing requires **only Python 3.12+** — no Kubernetes cluster, no Docker daemon, and no LLM API key, because the test suites are fully mocked. Cloning the repository and running `make setup` installs the workspace and runs six of the nine checks required to merge (the install smoke test and both Python 3.13 test jobs run only in CI). Active development happens in the `v4/` directory; `v1/`–`v3/` are frozen design-lineage snapshots that accept documentation fixes only. Any accepted contribution must preserve the human-in-the-loop approval gate on every mutating cluster operation. AI-assisted contributions are explicitly welcome when disclosed in the pull request.

- [GitHub repository](https://github.com/MSKazemi/kubeintellect): source code, issue tracker, and releases
- [Contributing guide](https://github.com/MSKazemi/kubeintellect/blob/main/CONTRIBUTING.md): dev setup, the exact CI gate commands, and a complete worked first pull request
- [Good first issues](https://github.com/MSKazemi/kubeintellect/contribute): curated small, well-scoped starting tasks for new contributors
- [AGENTS.md](https://github.com/MSKazemi/kubeintellect/blob/main/AGENTS.md): machine-readable repository rules for AI coding agents (agents.md format)
- [Discussions](https://github.com/MSKazemi/kubeintellect/discussions): questions, ideas, and Q&A with the maintainer
- [Governance](https://github.com/MSKazemi/kubeintellect/blob/main/GOVERNANCE.md): decision-making and the contributor ladder

## Blog

- [Blog](https://kubeintellect.com/blog): Technical writing on LLM agents, AI SRE, and Kubernetes operations.
- [KubeIntellect is peer-reviewed and open access](https://kubeintellect.com/blog/kubeintellect-peer-reviewed-paper): The framework is published open access in the Journal of Grid Computing (Springer, 2026) — 93% query resolution on a live cluster, +25 points over a raw GPT-4o baseline. The complete paper is free via DOI 10.1007/s10723-026-09837-6 (an earlier, shorter preprint is on arXiv 2509.02449).
- [How to choose an AI tool for Kubernetes](https://kubeintellect.com/blog/ai-kubernetes-tools-landscape): A capability-first guide — the difference between look, explain, reason, and act, and where to draw the safety line.
- [When an AI agent hits a wall, it should write its own tool](https://kubeintellect.com/blog/runtime-tool-synthesis): How KubeIntellect synthesizes, validates (AST checks, sandboxed test), and registers new Kubernetes tools at runtime, with 81.8% success on novel requests.

## Kubernetes Technical Guides

- [Kubernetes Debugging Guide](https://kubeintellect.com/kubernetes-debugging-guide): Systematic approach to diagnosing pod failures, node issues, network problems, DNS failures, storage failures, and resource exhaustion — with kubectl commands for every scenario
- [Fix CrashLoopBackOff](https://kubeintellect.com/fix-crashloopbackoff): Complete guide to diagnosing and resolving CrashLoopBackOff — covering exit codes, OOMKill, bad configuration, missing secrets, liveness probe misconfiguration, init container failures, and startup ordering problems
- [Kubernetes Root Cause Analysis](https://kubeintellect.com/kubernetes-root-cause-analysis): Structured RCA methodology for Kubernetes incidents — timeline reconstruction, blast radius mapping, change vector identification, signal correlation, postmortem templates, and common root cause patterns
- [Kubernetes AI Ops](https://kubeintellect.com/kubernetes-ai-ops): How AI and LLMs are transforming Kubernetes operations — anomaly detection, automated remediation, natural language cluster queries, capacity forecasting, and the AI Ops maturity model
- [Natural Language kubectl](https://kubeintellect.com/kubectl-natural-language): How AI translates plain-English intent into the correct kubectl command — the translation loop, cluster grounding, the read/write human-approval safety gate, worked English-to-kubectl examples, and the limits worth knowing
- [K8s Troubleshooting Playbook](https://kubeintellect.com/k8s-troubleshooting-playbook): Copy-paste kubectl playbooks for every common Kubernetes failure: Pending pods, ImagePullBackOff, Node NotReady, service routing failures, RBAC errors, stuck rollouts, HPA issues, DNS failures, PVC problems, Ingress misconfigurations, resource quota exhaustion, CronJob failures
- [AI SRE for Kubernetes](https://kubeintellect.com/ai-sre-for-kubernetes): What an AI SRE for Kubernetes is and how KubeIntellect delivers it — a human-governed operator that detects, diagnoses, and remediates with human approval and an audit trail
- [Kubernetes AI Security](https://kubeintellect.com/kubernetes-ai-security): Is it safe to let AI manage a Kubernetes cluster? KubeIntellect's security model — read-only by default, human-in-the-loop approval on every change, RBAC, protected namespaces, and Secrets the agent cannot read