Research

Peer-reviewed. Benchmarked. Open source.

KubeIntellect isn’t a prompt wrapper. Its architecture and evaluation were published in the Journal of Grid Computing (Springer). Here’s the paper, the results, the code, and how to cite it.

The paper

KubeIntellect: A Modular LLM-Orchestrated Agent Framework for End-to-End Kubernetes Management

Mohsen Seyedkazemi Ardebili, Andrea Bartolini

Journal of Grid Computing (Springer) · 2026 · Vol. 24, No. 3 · DOI 10.1007/s10723-026-09837-6

KubeIntellect is an LLM-powered system for end-to-end Kubernetes management through natural language, spanning read, write, delete, exec, access-control, and lifecycle operations via a supervisor coordinating 13 domain-specialized agents — with human-in-the-loop confirmation on every mutating action. When a request falls outside the static tool set, a Code Generator Agent synthesizes, validates, and registers a new tool at runtime. Evaluated on a live four-node cluster (170 pods across 18 namespaces), it resolved 93% of 200 operational queries and improved task success by 25 points over a tool-less GPT-4o baseline.

93%
Query resolution (186/200)
+25pts
Over a raw GPT-4o baseline
81.8%
Runtime tool synthesis (63/77)
7–10s
End-to-end per query
Read on Springer →Preprint (arXiv:2509.02449)

Changelog — built in the open

  1. July 2026Research

    Peer-reviewed and published

    • KubeIntellect was published in the Journal of Grid Computing (Springer), vol. 24, no. 3.
    • DOI: 10.1007/s10723-026-09837-6 — the framework and results are now citable.
  2. June 2026Discoverability

    Built for AI answer engines

    • Opened access for AI crawlers and added TechArticle structured data across guides.
    • A generative-engine-optimization (GEO) content pass so ChatGPT, Gemini, and Perplexity can cite KubeIntellect accurately.
    • Self-referencing canonicals and a cleaned-up metadata surface.
  3. May 2026Content

    Deeper guides and machine-readable docs

    • Expanded the Kubernetes troubleshooting guides to long-form, genuinely useful references.
    • Added an llms.txt manifest and pinned sitemap dates for stable indexing.
  4. April 2026Product

    Live demo, brand, and architecture docs

    • Shipped an interactive, in-browser demo of a real KubeIntellect session.
    • New brand mark and favicon; tightened the whole UI.
    • Documented the deployment workflow and system architecture for contributors.
  5. April 2026CLI

    kube_q command-line tool

    • Wired the kube_q CLI (v1.2.0) into the platform with proxied routing.
    • Runtime base-path handling so the CLI docs and app deploy cleanly.
  6. May 2025Milestone

    KubeIntellect started

    • First commit — the research prototype that became KubeIntellect.

Open source

The repositories

MSKazemi/kubeintellect

The KubeIntellect multi-agent system.

MSKazemi/kube_q

The kube_q command-line client.

Cite the research

Citation

KubeIntellect is peer-reviewed research, published in the Journal of Grid Computing. If it helps your work, please cite it:

@article{seyedkazemi2026kubeintellect,
  author    = {Seyedkazemi Ardebili, Mohsen and Bartolini, Andrea},
  title     = {KubeIntellect: A Modular LLM-Orchestrated Agent Framework
               for End-to-End Kubernetes Management},
  journal   = {Journal of Grid Computing},
  volume    = {24},
  number    = {3},
  pages     = {17},
  year      = {2026},
  publisher = {Springer},
  doi       = {10.1007/s10723-026-09837-6}
}

DOI: 10.1007/s10723-026-09837-6