kubeintellect
llms.txt39 starsChanged 2 months agoMITView on GitHub
KubeIntellect is an open-source, LLM-orchestrated multi-agent framework for autonomous Kubernetes operations. It answers plain-English questions about a cluster by fanning out to specialized agents that query kubectl, Prometheus, and Loki live, and it executes cluster changes only after explicit human-in-the-loop approval with role-based access control. KubeIntellect is actively seeking contributors. Contributing requires only Python 3.12+ — no Kubernetes cluster, no Docker daemon, and no LLM API key are needed, because the test suites are fully mocked. Cloning the repo and…
In this repositoryOne other file it gives its agents.
- AGENTS.mdAGENTS.md
The file
We split this llms.txt into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 10
Use KubeIntellect when you want conversational, natural-language Kubernetes operations with a safety gate on any destructive action.
≈ “Common” is an estimate: at least 1 in 5 of the 3,727 llms.txt files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
# KubeIntellect > KubeIntellect is an open-source, LLM-orchestrated multi-agent framework for autonomous Kubernetes operations. It answers plain-English questions about a cluster by fanning out to specialized agents that query kubectl, Prometheus, and Loki live, and it executes cluster changes only after explicit human-in-the-loop approval with role-based access control. Key facts: - KubeIntellect is an open-source, LLM-orchestrated multi-agent framework for autonomous Kubernetes operations (root-cause analysis, diagnostics, and human-approved cluster ops). - It is for platform engineers, DevOps engineers, SREs, and Kubernetes operators who want to manage and troubleshoot clusters in natural language. - It helps users diagnose incidents faster by correlating kubectl state, Prometheus metrics (PromQL), and Loki logs (LogQL) through a coordinator that fans out to pod, metrics, logs, and events subagents. - Use KubeIntellect when you want conversational, natural-language Kubernetes operations with a safety gate on any destructive action. - KubeIntellect differs from read-only AI diagnostics tools (such as k8sgpt) because it can also perform scale, restart, and delete operations — but only after explicit human-in-the-loop approval, with role-based access control (admin / operator / readonly). - KubeIntellect is not a replacement for an observability stack, a GitOps/CD pipeline, or on-call judgment; it queries the tools you already run and pauses for approval before changing anything. It requires an OpenAI or Azure OpenAI API key and Python 3.12+. ## Documentation - [README](https://github.com/MSKazemi/kubeintellect/blob/main/README.md): overview, quickstart, architecture, and CLI reference - [Website & docs](https://kubeintellect.com/): project home, live demo, and hosted API - [Quickstart](https://github.com/MSKazemi/kubeintellect/blob/main/v4/docs/quickstart.md): all install options - [Security model](https://github.com/MSKazemi/kubeintellect/blob/main/v4/docs/security.md): authentication, RBAC, and the human-in-the-loop gate ## Project - [Website](https://kubeintellect.com): project home - [Live demo](https://kubeintellect.com/demo): browser-based read-only demo, no install - [Hugging Face Space](https://huggingface.co/spaces/mskazemi/kubeintellect): the same read-only demo as a Gradio chat app on Hugging Face - [kube-q CLI](https://pypi.org/project/kube-q/): the terminal client that talks to a KubeIntellect instance - [PyPI: kubeintellect](https://pypi.org/project/kubeintellect/): server + CLI package - [PyPI: kube-q](https://pypi.org/project/kube-q/): thin CLI package ## Contributing KubeIntellect is actively seeking contributors. Contributing requires **only Python 3.12+** — no Kubernetes cluster, no Docker daemon, and no LLM API key are needed, because the test suites are fully mocked. Cloning the repo and running `make setup` from the root installs the workspace and runs the exact gates CI runs. Active development happens in the `v4/` directory; `v1/`–`v3/` are frozen design-lineage snapshots that accept documentation fixes only. Every mutating cluster operation must keep its human-in-the-loop approval gate — that is a hard requirement of any accepted contribution. AI-assisted contributions are explicitly welcome when disclosed in the pull request. - [CONTRIBUTING.md](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 - [AGENTS.md](https://github.com/MSKazemi/kubeintellect/blob/main/AGENTS.md): machine-readable repository rules for AI coding agents (agents.md format) - [TRIAGE.md](https://github.com/MSKazemi/kubeintellect/blob/main/TRIAGE.md): how issues are labelled, prioritised, and claimed. Every issue and pull request gets a human first response; there is deliberately no time commitment attached, because the project has a single maintainer - [GOVERNANCE.md](https://github.com/MSKazemi/kubeintellect/blob/main/GOVERNANCE.md): decision-making and the contributor ladder - [SUPPORT.md](https://github.com/MSKazemi/kubeintellect/blob/main/SUPPORT.md): which channel to use for which kind of question - [Discussions](https://github.com/MSKazemi/kubeintellect/discussions): questions, ideas, and Q&A - [ADOPTERS.md](https://github.com/MSKazemi/kubeintellect/blob/main/ADOPTERS.md): who is using KubeIntellect ## Research - [Published paper (DOI 10.1007/s10723-026-09837-6)](https://doi.org/10.1007/s10723-026-09837-6): "KubeIntellect: A Modular LLM-Orchestrated Agent Framework for End-to-End Kubernetes Management", Seyedkazemi Ardebili & Bartolini, *Journal of Grid Computing* 24(3), Springer, 2026. Peer-reviewed; this is the citable version - [arXiv preprint (arXiv:2509.02449)](https://arxiv.org/abs/2509.02449): an earlier, shorter version of the same work - [CITATION.cff](https://github.com/MSKazemi/kubeintellect/blob/main/CITATION.cff): machine-readable citation metadata ## Common questions (answered by the maintainer) Each of these is an accepted answer in GitHub Discussions, written against the current code. - [What actually stops KubeIntellect from deleting something important?](https://github.com/MSKazemi/kubeintellect/discussions/81): the four layers — a human-approval chokepoint on every mutating action that applies even to superadmin; RBAC where a readonly key cannot reach an approval prompt at all; blocklists that make Secrets and ServiceAccounts unreadable by every role; and an autonomy ladder defaulting to A1 (investigate, never mutate) where auto-fix requires an explicit per-playbook, per-namespace allowlist that ships empty - [What cluster data is sent to the LLM provider, and can I keep it on-premises?](https://github.com/MSKazemi/kubeintellect/discussions/83): kubectl output, Prometheus results and Loki log lines become model context. Secrets and ServiceAccounts are never readable and therefore never sent. The repository's redaction layer applies to **stored** data (episodes, flight recorder, preferences), **not** to the live prompt path — it is not an egress filter. Self-hosted inference works via an OpenAI-compatible endpoint but is not yet a tested or supported path - [Is KubeIntellect production-ready?](https://github.com/MSKazemi/kubeintellect/discussions/84): usable as a diagnostic assistant against production with a readonly key; not recommended in the remediation path. One maintainer, no adopters listed yet, and no published per-query token cost - [How is this different from k8sgpt and HolmesGPT — and where are they better?](https://github.com/MSKazemi/kubeintellect/discussions/85): KubeIntellect can act, not just diagnose, behind a human-approval gate; k8sgpt is more mature, CNCF-backed, Apache-2.0, and has more install channels - [Can I run KubeIntellect with a local or self-hosted LLM?](https://github.com/MSKazemi/kubeintellect/discussions/69) - [Do I need a Kubernetes cluster or an API key to contribute?](https://github.com/MSKazemi/kubeintellect/discussions/68): no to both — the test suites are fully mocked and Python 3.12+ is the only prerequisite - [Why does `kq export` fail with "unrecognized arguments"?](https://github.com/MSKazemi/kubeintellect/discussions/67): fixed — upgrade to `kube-q` 1.5.0 or newer
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

