# AI SDLC

- Page: https://hub.theagilemonkeys.com/topics/ai-sdlc
- Updated: 2026-11-03

The AI SDLC, also called the agentic SDLC, is a software delivery lifecycle rebuilt around AI agents. Agents turn an issue into a plan, implement the change, run the checks and open the pull request. People decide what to build, review the plan and the running change, and decide what ships. Agentic engineering is how one engineer works with agents; the AI SDLC is the delivery system a team builds around them.

Ours runs on Facility, our open source platform for Claude Code and Codex. Each task keeps its conversation, environment and a record of every change, and GitHub reviews and required checks decide what merges.

On one internal project, 27 of 30 completed agent pull requests merged after human review. 57% merged with no change requests and no human fixup commits. The median from issue to merge was 22 hours.

That is one project and a small team, and Facility is early software. In regulated companies the harder part is making context, controls and evidence usable by agents.

## See it working

- [AI SDLC: Building a Software Delivery System Around Claude](https://hub.theagilemonkeys.com/videos/ai-sdlc-building-a-delivery-system-around-claude) (2026-09-29): Talk from the Claude Community House in Barcelona (21 Sep 2026): how Facility, our open source AI SDLC, works. Agents propose and implement; humans approve and merge. In English.
- [Claude Community House, Barcelona](https://hub.theagilemonkeys.com/events/claude-community-house-barcelona-2026-09-21) (2026-09-21): We gave the AI SDLC talk on how we built a software delivery system around Claude, with Flavia Marsano, Adrián Lorenzo and Álvaro López Espinosa.

## What we measured

- [90% of agent pull requests merged after review](https://hub.theagilemonkeys.com/findings/finding-ai-sdlc-90-percent-tickets-merged-or-closed) (2026-09-21): On one internal project, 27 of 30 completed agent pull requests merged after human review. 57% merged with no change requests and no human fixup commits, and the median from issue to merge was 22 hour
- [Our Take on AI SDLC ROI in Regulated Industries](https://labs.theagilemonkeys.com/posts/ai-sdlc-roi-regulated-industries-enterprise-legible-to-agents/) (2026-06-09): Why regulated companies should focus less on baseline agent capability and more on the operating environment that makes enterprise work usable, governed, and measurable.

## The open source behind it

Start with one repository.

- [Facility](https://hub.theagilemonkeys.com/open-source/repo-facility) (2026-10-01): Scalable, observable and governed AI SDLC platform.
- [Claude Dev Kit](https://hub.theagilemonkeys.com/open-source/repo-claude-dev-kit) (2026-09-28): Open source, stack-agnostic issue-to-PR workflow plugin for Claude Code: quality gates, tracker and PR-host adapters.