The AI-native SDLC playbook
Anthropic
Last updated October 13, 2026
The AI-native SDLC playbook is a free, text-based course from Anthropic for engineering, platform, and security leads at organizations that have adopted Claude Code. It lays out twelve plays, spread across the stages of the software development lifecycle (SDLC), that use Claude to speed up the work surrounding the code: approvals, reviews, handoffs, and policies. The plays come from the practices of Anthropic's Applied AI team. Among them: recording a request in intent.md, a brief file kept under version control that says what is needed, the reason, and the limits, then working it up with Claude into a requirements and design spec in a single session; choosing between CLAUDE.md, a skill, or a hook as the home for a piece of team knowledge; running the agent's evaluations again in continuous integration (CI) each time its setup changes; having Claude review every pull request while people keep reviewing some code themselves; using hooks to pause steps such as a release until someone signs off; and treating a production metric that drifts out of its normal range as a fresh intent.md, which starts planning again. It takes about an hour to read across 14 lessons, and Anthropic also publishes the whole playbook as one page on the Claude blog.
What you'll learn
- How each of the six SDLC stages changes in an AI-native process, and picking a first play using the dependency graph
- Recording a request in intent.md and working it into a requirements and design spec with Claude
- Deciding where a piece of team knowledge should live: CLAUDE.md, a skill, or a hook
- Running parallel sessions and subagents, and giving Claude a feedback loop
- Rerunning evaluations in CI whenever the agent's setup changes, and bringing Claude into pull request review
- Using hooks as approval gates that pause a release or similar step until someone signs off
- Treating a production metric that drifts out of range as a new intent.md
Frequently asked questions about The AI-native SDLC playbook
Who is The AI-native SDLC playbook for?
Engineering, platform, and security leads in large organizations, especially regulated ones, whose teams use Claude Code but whose reviews, approvals, and handoffs have not sped up to match.
Is The AI-native SDLC playbook free?
Yes — The AI-native SDLC playbook is free to take. Running the plays needs Claude Code, plus a Git repository and a continuous integration (CI) pipeline that you have permission to change.
What are the prerequisites for The AI-native SDLC playbook?
You should use Claude Code comfortably in your daily work and have a Git repository plus a continuous integration (CI) pipeline that you have permission to change. Individual plays list their own prerequisites, and some list none.
Do you need to code for The AI-native SDLC playbook?
Yes — The AI-native SDLC playbook assumes you can already code, though the course itself is mostly theory.
Why we suggest this course
Worth reading if your team already writes code with Claude Code but releases still wait on the same reviews, handoffs, and approvals as before. The playbook works through that problem stage by stage, and it is written for large and regulated enterprises, so it keeps human approval where a change process requires it rather than removing review. You can start with a single play: the first lesson maps how the plays depend on each other. One thing to know: the plays expect you to have authority over your organization's planning, review, testing, and release processes, not just its coding practices.