Agents propose. Hooks block. Git decides.

The engineering guide to bringing Claude Code, MCP and subagents into your daily Terraform, Kubernetes, AWS and CI/CD work, with the safeguards a production environment demands.

The book is free with sign-up. The interactive guide costs $19 for 1 year and is updated every week.

The explorer shows a sample of what ends up in your repository. There are 52 templates (subagents, hooks, rules, skills and CLAUDE.md), each one explained in the chapter where it is built.

your-repo/ Click a file
  • .claude/
  • agents/
  • hooks/
  • rules/
  • skills/


        
Cover of The AI-Native DevOps Engineer, edition 1.1.2

The whole guide is now a book, and the book is free

The Official Guide to DevOps with AI, free in PDF and EPUB. Edition 1.1.2 has all 16 chapters as they stood on October 26, 2026. Sign up and get the links by email. The guide keeps changing every week in the app, and that is what the $19 subscription covers.

Book, freeFrozen on a date. A new edition every 3 months.
Interactive guide, $19 for 1 yearAlways the latest version, with search, interactive examples and a changelog.

One system, in the order it runs in production

Installing Claude Code takes five minutes. What the guide teaches is the chain that keeps an agent from doing damage: each step delivers something the next one uses.

Illustrative example: a nodepool change moving through all six steps. Click a step or run the whole thing.


      

This guide has a changelog

Claude Code, MCP and the providers change every month. A static PDF is outdated the day it ships. Here every change is published with its date, section and what was added, and your access includes all of them for a year.

Last updated: 11/16/2026

See the full history
  1. Provider upgrades with acceptance criteria (section 5.9.8)

    A scheduled agent upgrades one Terraform provider a week, adapts the code to the upgrade guide and only opens the PR once a rubric approves the evidence: complete lock, clean validate and tflint, no replacement without a reason.

  2. Continuous upgrade readiness (section 7.13.10)

    A scheduled agent runs as a pod in the cluster itself, with read-only RBAC, and tells you every week whether it is ready for the next minor: kubelets, CRDs, PDBs, webhooks, add-ons and who still calls a deprecated API.

  3. Scheduled drift with triage (section 5.9.6)

    A daily read-only plan in every environment, on a self-hosted worker with the plan-only role: the agent remembers what it already reported, honors exceptions approved in git and keeps one issue per environment.

  4. Scheduled agents as code (section 4.7)

    Managed Agents running on cron with ant apply, a spend cap per run and a plan on every PR, plus the self-hosted sandbox that keeps tool execution inside your network.

Written for the people accountable for the environment

The guide assumes you already run something that can't go down. The question isn't whether AI can write Terraform. It's what you let it execute, and behind which safeguard.

It's not for anyone looking for NoOps. The guide argues for human-in-the-loop, governance and technical accountability. If the plan is to let an agent apply straight to production, this material will disagree with you the whole way.

DevOps, SRE and Platform Engineers

Who already maintain Terraform, Kubernetes and pipelines and want speed without giving up review and rollback.

Tech leads and Staff engineers bringing AI to their team

And who need ADRs, rules and permissions to show leadership there is control, not just enthusiasm.

People moving into DevOps

The chapters explain why each safeguard exists, not just the command. That cuts years of trial and error.

16 chapters, 148 sections, 700+ pages

Every chapter is in the free book. In the interactive guide, they get new sections and revisions every week.

01The system the guide builds
09Observability and incidents
02AI fundamentalsLLMs, context, MCP
10Container and Kubernetes securitySpec-driven, executor gated
03The modern IDEVS Code and Claude Code
11FinOps with AITagging, Infracost, costs via MCP
04AI agentsSubagents, ADRs, spec-driven
12Runbook RAG
05Terraform with AI
13Guardrails and professional useFail-closed hooks, OpenTelemetry audit
06AWS with AIAWS MCP, Bedrock, spec-driven
14AI-assisted GitOpsArgoCD, ApplicationSet, Argo Rollouts
07Kubernetes with AIK8sGPT, MCP, EKS upgrades with Karpenter
15Governance and adoption
08CI/CD with AIOIDC, GitOps, spec-driven
16Conclusion and the future

From readers

"Finally the first AI guide focused on DevOps that covers what matters. No fluff."
Leonardo Magalhaes, Tech Lead
"I didn't know several of the tools mentioned... MCP, RAG, all together. Now I never leave the IDE."
Claudiano Silva, Senior SRE
"Highly technical and conceptual at the same time. It gave me the foundation to discuss strategy with leadership."
Bruno Sereballos, Senior DevOps Engineer
"Even with little experience, I closed the gap with other professionals and was able to apply many of the concepts."
Bernando Gomes, Junior DevOps
Filipe Motta

Who wrote it

Filipe Motta. Senior DevOps, Cloud and Platform Engineer since 2015.

This guide was written from real problems: incidents, automation that failed and hard decisions in critical production, cloud, Kubernetes and automation at scale. Every safeguard in it exists because, at some point, someone needed it.

CKA CKAD CKS AWS Solutions Architect AWS DevOps Engineer GCP DevOps Engineer Terraform Associate LPIC-1

More about me at filipemotta.me and on LinkedIn.

Free book or interactive guide

The content is the same. The difference is when it reaches you and how you read it.

Book

Edition 1.1.2, October 26, 2026

$0

Free, with name and email sign-up.

  • All 16 chapters, as they stood on the edition date
  • PDF and EPUB, to read on any device
  • The file is yours, forever
  • An email when the next edition ships, every 3 months
Download the free book

Prefer paper? The print edition is coming soon on Amazon (paperback and Kindle), priced close to the printing cost. Learn more.

Frequently asked questions

Does it work for Azure and GCP, or only AWS?
The concepts (Terraform, Kubernetes, MCP, agents, hooks) are cloud-agnostic and work on Azure, GCP and on-prem. Chapter 6 is dedicated to AWS, and there are public skills for GKE too.
Is it practical or just theory?
Practical, with production context, trade-offs and clear limits. The templates you saw at the top of this page are the same ones the guide explains line by line.
Does the guide keep up with changes in AI tools?
The interactive guide does, every week. You can check before subscribing: the release notes are public and list every section added, with dates. The book picks up those changes in each new edition.
How long do I have access?
For the interactive guide, 1 year, including every update in that period. One-time payment of $19, no automatic renewal. The book is yours forever.
Does the free book have the same content as the guide?
It has the same 16 chapters, frozen on the edition date (1.1.2 is from October 26, 2026). Anything published after that reaches the interactive guide first and goes into the book in the next edition. Download it here.
So why pay $19?
Because Claude Code, MCP and infrastructure tooling change every week. The interactive guide follows those changes the day they ship, and the reader has search, tabs and interactive examples that a PDF cannot have. If one edition every few months works for you, the book is enough.
I already bought the guide. What changes for me?
Nothing about your access: it stays valid for the period you purchased, with every update. You can also download the book, like anyone else.
I'm new to DevOps. Can I keep up?
Yes. The guide explains the reasoning behind each decision, and early-career readers say it narrowed the gap with more experienced professionals.

Start with the free book. Subscribe when you want the updates.