Amazon is sunsetting its Q Developer AI coding assistant and replacing it with Kiro, a radically different approach to AI-assisted development that requires writing detailed specifications before code generation begins. The company announced that new signups for Q Developer closed on May 15, 2026, with full end-of-support coming April 30, 2027. This marks a significant philosophical shift in how Amazon believes developers should interact with AI tools.

Why Is Amazon Abandoning Q Developer for Something Completely Different?

When Amazon launched Q Developer, the tool followed the same playbook as most AI coding assistants: you describe what you want, the AI generates code, and you iterate through chat. But after a year of real-world usage across VS Code, JetBrains, Eclipse, and Visual Studio, Amazon concluded that this approach was fundamentally limited. The company determined that developers need AI that understands entire projects, not just individual code snippets.

Kiro represents a complete departure from this model. Instead of reacting to prompts, Kiro operates on structured specifications that drive implementation across your entire codebase. Before writing a single line of code, Kiro's agent produces three formal documents: requirements.md with user stories and acceptance criteria, design.md with system architecture and component breakdowns, and tasks.md with a numbered implementation checklist. The agent then executes these tasks while the developer reviews and approves each artifact before proceeding.

This is not a cosmetic difference from tools like Cursor or Claude Code. It is architecturally different, meaning teams must do software design before software development. For regulated industries or complex projects where structured documentation is already mandatory, this discipline restoration makes sense. For solo developers shipping fast, it introduces friction with uncertain payoff.

What Are the Key Capabilities That Make Kiro Different?

• Specs: Developers define what they want to build in structured, natural-language requirements that Kiro uses to drive implementation end to end.
• Hooks: Automated triggers that run on file save, commit, or other events to enforce standards, run tests, or update documentation without manual intervention.
• Steering files: Project-level configuration that gives Kiro persistent context about your architecture, conventions, and constraints.
• Custom subagents: Specialized AI agents you define for domain-specific tasks like security review, API contract validation, or infrastructure provisioning.
• Powers: Composable capability modules that extend Kiro's agentic behavior for your specific workflows.

The agent hooks feature is particularly novel. These are event-driven automations that fire without manual prompting; save a React component and the test file updates automatically, modify an API endpoint and the README refreshes, or trigger a pre-commit and a security scan runs. Kiro also supports Figma MCP (Model Context Protocol) integration so design validation fires when you modify HTML or CSS. This is closer to CI/CD automation inside the IDE than it is to autocomplete, making Kiro more compelling for teams than for individuals.

How to Migrate From Q Developer to Kiro

• Timeline for action: Existing Q Developer Pro subscriptions retain access until April 30, 2027, but new signups have been blocked since May 15, 2026, and the latest models including Opus 4.7 are now Kiro-exclusive.
• Model availability: Starting May 29, 2026, Opus 4.6 is no longer available on Q Developer Pro, while Opus 4.5 and other existing models remain available; staying on Q Developer means accepting a capability freeze for the next eleven months.
• Pricing considerations: Q Developer Pro provided roughly 1,000 requests per month at $20, while Kiro Pro provides approximately 350 mixed requests at the same price, with spec-mode requests costing $0.20 per credit and vibe-mode requests costing $0.04, representing a 5x premium for using Kiro's signature feature.
• Transition support: AWS is offering a $20 first-upgrade credit to soften the transition, and existing Q Developer subscriptions can continue adding new users until the April 2027 deadline.

The pricing math reveals an important trade-off. Q Developer Pro gave developers roughly 1,000 requests per month at $20, while Kiro Pro gives approximately 350 mixed requests at the same price. That is a real reduction in volume, and the credit system adds cost complexity that Q Developer never had. Spec-mode requests cost $0.20 per credit, while vibe-mode requests cost $0.04, creating a 5x premium for using Kiro's signature feature over its basic chat mode.

AWS argues that structured specs reduce downstream logic errors by 23 to 37 percent. That is a plausible claim, since formalizing requirements before coding is standard practice in software engineering for a reason. However, this is a metric without public methodology, and teams would need to verify this against their own codebases before justifying the premium at scale.

What Does This Mean for the Broader AI Coding Tool Market?

The agentic IDE market is consolidating, not expanding. GitHub Copilot suspended new sign-ups recently, and now Amazon Q is being sunset. The field is shaking out around a smaller set of serious players: Cursor at the speed end, Claude Code at the reasoning end, and Kiro staking out a spec-and-structure position in the middle.

Most experienced developers are not picking one tool anymore. The common pattern in 2026 is using Kiro for structured feature planning on complex projects, Cursor for rapid iteration, and Claude Code for deep architectural reasoning. This is not idealism; it is what happens when different tools have genuinely different strengths rather than marginal differences.

For Q Developer users, the immediate action is straightforward: Kiro is the path AWS has chosen, model access is better there, and the $20 first-upgrade credit softens the transition. But whether developers lean into spec-driven development or treat Kiro mostly as a vibe-mode replacement for Q Developer is a decision about how their team actually works, and the IDE cannot answer that question for them.