AWS made it official last week: Amazon Q Developer, the company’s AI coding assistant, is on its way out. New signups are blocked from May 15, 2026. IDE plugins and paid subscriptions reach full end-of-support on April 30, 2027. The replacement is Kiro, an agentic IDE that AWS has been building in parallel.

For the tens of thousands of enterprise teams that adopted Q Developer over the past two years, this is a forced migration. But looking at what Kiro actually does, it is less of an upgrade and more of a rethink of how AI fits into software development.

What Changes With Kiro

Amazon Q Developer worked like most AI coding assistants: you ask, it suggests, you accept or reject. Kiro takes a different approach. AWS calls it “spec-driven development” — instead of responding to individual prompts, Kiro works from structured specifications that define what you are building before a single line of code is written.

When you describe a feature in Kiro, it creates three files automatically: a requirements document, a design document, and a task breakdown. These become living artifacts that update as the build progresses. The system asks clarifying questions rather than making assumptions. It is designed to catch design problems before they become code problems — which is exactly where vibe coding falls apart.

A few features stand out for enterprise teams specifically:

Steering files let you encode your architecture conventions, library choices, and coding standards into persistent project-level configuration. Every Kiro agent respects these rules without you needing to repeat them per session. Teams can distribute standard steering files via MDM or Group Policies, giving organizations control over how AI behaves across all developer machines.

Hooks are automated triggers that fire on file saves, commits, or other events. You can configure hooks to run tests automatically, flag security issues, or update documentation when related code changes. This replaces a whole category of manual review steps that slow down shipping.

Custom subagents let you define specialized AI agents for domain-specific tasks: security review, API contract validation, infrastructure provisioning. Rather than a single generalist assistant, teams build a tailored set of agents that match their actual workflow.

Kiro is built on Code OSS, so VS Code settings and Open VSX extensions carry over. The underlying models run on Amazon Bedrock, and Kiro is the only AWS tool that now offers Claude Opus 4.7 — Q Developer Pro is capped at Opus 4.6 as of May 29, 2026.

The Timeline

The transition is staggered to give organizations time to move:

• May 15, 2026: New Q Developer Free Tier and Pro subscriptions blocked
• May 29, 2026: Opus 4.6 removed from Q Developer Pro; Opus 4.7 available only on Kiro
• April 30, 2027: Full end-of-support for Q Developer IDE plugins and paid subscriptions

Existing Q Developer Pro subscribers can keep adding users until the 2027 deadline. The AWS Management Console version of Q Developer and integrations with Slack and Teams are not affected by this sunset.

Kiro pricing starts free at 50 agentic interactions per month. Pro is $19 per month for 1,000 interactions. Pro+ is $39 per month for 3,000. Enterprise plans are available through AWS directly.

What This Means for Business

The move signals something broader than a product refresh. AWS is betting that the “chat and accept” pattern of first-generation AI coding assistants is not how enterprise software actually gets built at scale. Vibe coding works for prototypes. It breaks down when you need consistent architecture, audit trails, documented decisions, and maintainable codebases.

Kiro’s design reflects lessons from two years of enterprise AI coding adoption. The patterns that caused the most problems — AI agents that ignored established conventions, generated undocumented changes, required constant supervision — are exactly what steering files, specs, and hooks are designed to prevent.

For data teams specifically, the spec-driven approach is a better fit for the kind of analytical work that requires upfront design: data pipeline architecture, model deployment workflows, reporting infrastructure. Prompting an AI to build a Power BI data model without a spec is how you end up with something that works once and breaks every time the source changes.

If your team is on Q Developer today, the May 15 deadline for new signups is a forcing function. Teams adding developers after that date will go onto Kiro by default. The practical path forward is to run both in parallel through mid-2026, then consolidate to Kiro before the 2027 cutoff.

The bigger question is what this means for the other AI coding assistants — Cursor, Windsurf, GitHub Copilot — that built their businesses on the same chat-and-suggest pattern that AWS is now walking away from. Spec-driven development is a bet that the next two years of enterprise AI adoption look very different from the last two.