Amazon launches Kiro, its own Claude-powered challenger to Windsurf and Codex

Amid the big news that Windsurf is being acquired by Cognition (after its founders went to Google), developers interested in AI-powered coding may be on the hunt for new alternatives.

In a bit of fortuitous timing, today also saw Amazon’s release of Kiro, a new agentic integrated development environment (IDE) built to help developers move from prototype to production using AI workflows grounded in structure, planning and engineering rigor.

Kiro uses Claude Sonnet 3.7 and 4.0 as the default model backends. Users can switch between them, and future support for other models may be added.

Now in public preview, Kiro runs on macOS (Intel and Apple Silicon), Windows and Linux for free to start (limited to 50 interactions per user per month), with additional pricing tiers starting at $19 for more features.

Why should developers check out Kiro?

Kiro aims to bridge the gap between “vibe coding” — allowing AI to generate full blocks of code or entire software processes and applications from plain text instructions, typically for rapid prototyping and iteration — and the more demanding process of delivering secure, maintainable and scalable applications in real-world environments.

The tool combines AI agents with project specifications, technical architecture and automated task management to support a complete software development lifecycle inside a single interface.

Kiro vs. Q Developer?

But didn’t Amazon already have its own AI-code completion tool, Q Developer? Yes, and that’s still available.

So why launch a whole new product and brand name that offers some of the same functionality? Sources at Amazon told VentureBeat that “Kiro is a general-purpose agentic IDE for developers to work with any platform of their choice,” as opposed to Q Developer, which is more limited in its support for third-party IDEs, restricted to VSCode, JetBrains, Eclipse and Visual Studio.

In addition, the sources pointed out that Kiro’s agentic spec-driven development was radically different from the code suggestions offered on discrete snippets by Q Developer. They said some developers may even prefer to use both in tandem, which is supported via the Q Developer Pro subscription which starts at $19 per month per user.

From prompt to production with spec-driven development

Kiro’s key differentiator is its spec-driven development model, which guides the process from ideation to implementation. A simple prompt like “add a review system” triggers a chain of AI-assisted outputs that include:

User stories with acceptance criteria in EARS (Easy Approach to Requirements Syntax) format;

Design documents with data flow diagrams, TypeScript interfaces and API schemas;

Task lists and sub-tasks automatically sequenced by dependency, with tests, loading states and accessibility built-in.

Developers can execute these tasks one at a time through Kiro’s built-in agent interface, with inline diffs, progress tracking and access to historical agent execution logs. As development proceeds, Kiro keeps specs in sync with the codebase, helping teams avoid the typical drift between documentation and implementation.

Agent hooks automate routine quality tasks

Kiro’s agent hooks allow developers to configure automation triggers for everyday tasks like regenerating tests, updating documentation or running security scans.

Hooks can be tied to actions such as saving files, editing components or pushing commits. Once set up and checked into the repo, they provide team-wide consistency in code quality and standards enforcement.

For instance, developers can define a hook to ensure new React components follow the ‘Single Responsibility Principle’ or trigger a secrets scan before commits. This approach adds automated quality control without slowing down individual developers.

Kiro is built on Code OSS, the open-source foundation of Visual Studio Code maintained by Microsoft. It provides the core editor experience without proprietary services, allowing third parties like Kiro to build their own IDEs with full compatibility with VS Code extensions and settings.

As such, Kiro remains compatible with VS Code extensions, settings and UI conventions. It also supports:

Model Context Protocol (MCP) for connecting external tools;

Agentic multi-modal chat, using files, URLs or documents as context;

Steering rules to customize and constrain agent behavior across a codebase;

Social login via GitHub or Google, with no AWS account required.

Pricing and availability

Kiro is currently free for all users during its preview period, including Amazon Q Developer and Q Developer Pro subscribers.

Preview access includes “generous” usage limits aimed at letting developers explore Kiro without frequent disruptions.

After the preview period ends, users will have a choice of three subscription tiers:

PlanMonthly priceIncluded agentic interactionsNotesKiro Free$050 per monthSpecs, hooks, steering and MCP support includedKiro Pro$191,000 per monthAll features from Free, plus higher usage quotaKiro Pro+$393,000 per monthDesigned for heavy users or teams

Agentic interactions include any direct invocation of Kiro agents — such as initiating a spec, triggering a hook or sending a chat prompt. The subsequent processing work (like multi-step task execution) does not count toward the quota.

Users on paid plans will also be able to purchase additional interactions at $0.04 each, but overage billing must be explicitly enabled.

The AI-assisted development ecosystem is becoming increasingly crowded, with several prominent IDEs and agents competing for developer attention. Here’s how Kiro stacks up:

ToolSummaryPricingAmazon Q DeveloperMulti-environment AI assistant integrated into AWS, IDEs, CLI and chat. Agentic workflows via terminal or IDE. Great for cloud ops, migrations and automation. Free and Pro tiers available.Free; Pro at $19/user/monthClaude Code (Anthropic)CLI-first coding assistant with chat-based iteration, plan/edit modes and diff views. Strong for interactive code development, less structured than Kiro.Free; Pro at $17/month or $20 billed monthlyGitHub Copilot (Microsoft)Inline code completion tool in VS Code, GitHub. Best for quick assistance. Lacks structured planning or workflow support.Free trial; Pro $10/month or $100/year; Pro+ $39/monthCursorVS Code–based AI editor for conversational edits and navigation. Optimized for solo coding, minimal planning support.Free; Pro $20/month ($16/month yearly)Windsurf (OpenAI, acquired)Discontinued AI IDE focused on rapid code editing. Offered minimal planning. Key staff/IP acquired by Google and Cognition in July 2025.Free; Pro $15/month; Teams $30/user/month Cognition / DevinMulti-agent system capable of autonomous software engineering from planning to deployment. Developer-in‑loop model.Start at $20/month; older tiers at $500/monthKiroPlanning-first AI IDE with structured artifacts like specs, design docs and task trees; supports feature planning, implementation and QA automation. Developer‑in‑loop.Free (50 interactions/mo); Pro: $19/user/mo (1,000 interactions); Pro+: $39/user/mo (3,000 interactions)

Showcasing Kiro in action: ‘Spirit of Kiro’

To demonstrate Kiro’s capabilities in a real-world context, Amazon released a full demo project called “Spirit of Kiro”, an open-source crafting game. The project serves as a hands-on example of how Kiro can be used throughout the development lifecycle.

Over 95% of the game’s codebase was generated by prompting Kiro. The game features unique, procedurally generated items with customizable properties like damage, quirks and enchantments. Players can combine, break down and sell items — providing a complex system that highlights Kiro’s strengths in managing interconnected components and evolving feature sets.

The repo includes:

Architecture documentation;

App security overviews;

A deliberately incomplete sample branch with bugs (for testing spec workflows);

A roadmap with feature ideas for future contributions.

The project is also intended as a learning resource. Files like CHALLENGE.md, architecture.md and guiding-principles.md are designed to walk developers through Kiro’s specs, hooks and agentic workflows in practice.

Developers can clone, run or deploy the project locally or on AWS infrastructure, and open-source contributions are welcome via GitHub.

Developer response and early impressions

Kiro’s launch generated active discussion on startup accelerator Y Combinator’s popular developer forum Hacker News, where Nathan Peck, senior developer advocate for generative AI at AWS (username NathanKP) offered technical context and responded to questions.

He emphasized that Kiro reflects Amazon’s internal engineering practices and is designed to help developers scale from small ideas to robust, production-ready systems.

Initial community reactions were mixed, but developers were intrigued, praising the emphasis on specs, hooks and structure.

Some compared it favorably to tools like Claude Code and Cursor, citing the improved rigor in building and documenting features. Others voiced concern over tool churn and switching costs, while some preferred command line interface (CLI)-based tools or simpler interfaces.

Feedback also surfaced around authentication bugs, platform compatibility and desire for dev container support. These early responses reflect both curiosity and the high expectations developers now have for AI coding tools.

Kiro enters a crowded field but appears to carve out a niche with its structured, spec-first philosophy and support for developer-in-the-loop workflows.

It’s not trying to replace developers or automate entire codebases blindly. Instead, it’s offering a more disciplined way to collaborate with AI from planning to delivery.

With its preview now open and pricing models outlined, Kiro may appeal most to teams and individuals looking to build not just faster, but more thoughtfully — with long-term maintainability, clarity and quality built in.