What Kiro Does
Kiro arrives at a moment when the AI coding tool landscape is saturated with assistants that help you write code faster but do very little to ensure that code is actually production-ready. Built by a small team within AWS and launched in public preview at the AWS Summit New York in July 2025, Kiro takes a fundamentally different approach: instead of optimizing for speed of code generation, it optimizes for the quality and structure of the entire development process. The result is an IDE that feels less like an autocomplete engine and more like a junior developer who insists on writing a spec before touching a single file.
Spec-Driven Development
The spec-driven development workflow is Kiro's defining innovation. When you start a new feature, Kiro takes your natural language prompt and transforms it through a three-phase process: Requirements, Design, and Tasks. Requirements are expressed in EARS notation — a structured format that makes acceptance criteria explicit, including edge cases that developers typically handle manually. The Design phase produces architecture decisions and system design documentation. Tasks break the implementation into discrete, dependency-ordered units. Only after you have reviewed and approved these artifacts does Kiro begin writing code. This structured approach directly addresses the biggest problem with what the community calls vibe coding — building software through intuitive prompting without any documented plan.
VS Code Foundation and Hooks
Kiro is built on Code OSS, the open-source foundation of Visual Studio Code, which means it is immediately familiar to the vast majority of developers. Your existing VS Code extensions, settings, and keyboard shortcuts carry over. The free tier includes access to open-weight models and Claude Sonnet 4.5, while paid plans add premium-model access. Unlike Amazon Q Developer, which is deeply integrated into the AWS ecosystem, Kiro is deliberately positioned as cloud-agnostic — you do not need an AWS account to use it, and you can sign in with Google or GitHub. This separation is strategic: AWS wants Kiro to appeal to developers regardless of which cloud platform they use.
The Hooks system is one of Kiro's most practical features. Hooks are event-driven automations that trigger AI agents to perform tasks in the background as you work. You can set up a hook that automatically generates documentation whenever a file is saved, runs code quality checks on every commit, or updates test files when implementation code changes. This is conceptually similar to Git hooks or CI triggers, but powered by AI agents that understand the semantic meaning of your changes rather than just running static scripts. For teams that struggle with documentation debt and test coverage gaps, hooks provide an automated safety net.
MCP Support and Practical Use
MCP (Model Context Protocol) support extends Kiro's capabilities beyond what the built-in AI models know. You can connect MCP servers for AWS documentation, internal knowledge bases, proprietary APIs, and specialized tools — all while maintaining security boundaries. The AWS Documentation MCP server is particularly useful, giving Kiro access to current service documentation and API references rather than relying on potentially outdated training data. For enterprise teams, the ability to connect private knowledge bases through MCP means Kiro can reference internal coding standards, architecture decision records, and company-specific documentation without sending that data to external services.
In practice, Kiro works well for structured, planned development but requires patience. Creating a full project spec can consume a meaningful number of credits, which matters because the free tier provides 50 credits per month. The spec workflow excels at complex features that benefit from upfront planning — database schema design, API architecture, multi-service integrations — but feels like unnecessary overhead for quick fixes or simple utility functions. Kiro includes a Vibe mode for those situations, offering standard chat-based coding assistance similar to Cursor or Copilot, but this is clearly not where the tool's strength lies.
Pricing and Privacy
Pricing is competitive but credit-limited. The free tier offers 50 credits per month, which is enough to evaluate the tool but not enough for daily use. The Pro plan is listed at $20 per month for 1,000 credits, Pro+ at $40 per month for 2,000 credits, and Power at $200 per month for 10,000 credits, with pay-per-use overage on paid plans. Compared with flat-seat coding tools, Kiro's credit model can feel restrictive during intensive development sessions, so teams should estimate monthly credit usage before committing.
The privacy model benefits from AWS's enterprise infrastructure heritage. Code is processed through AWS's secure infrastructure, and Kiro does not store or use your code for model training. The MCP architecture adds an additional layer of data control by allowing you to keep sensitive information within your own MCP servers rather than sending it through the AI model. For organizations in regulated industries, this hybrid approach — cloud AI processing with on-premises knowledge management — provides a reasonable balance between capability and compliance.
Limitations
Kiro's current limitations are real but expected for a preview product. Spec documents are mostly static — they do not automatically update as code evolves, which means specifications can drift out of sync with implementation over time. The AI sometimes struggles with complex infrastructure tasks like database connection setup, occasionally generating documentation instead of solving the actual problem. The tool is also not yet available as a CLI-only experience for developers who prefer terminal-based workflows, though a Kiro CLI was announced. Extension compatibility is limited to Open VSX rather than the full VS Code Marketplace, which means some popular extensions are unavailable.
The Bottom Line
Kiro represents AWS's bet that the next phase of AI-assisted development is not about generating code faster but about generating better, more maintainable software from the start. For teams that have experienced the technical debt accumulation that comes with unstructured AI coding, Kiro's spec-driven approach offers a compelling alternative. It is not the fastest AI coding tool, and it is not the most flexible, but it may be the most disciplined — and for production software, discipline often matters more than speed. The preview status means early adopters should expect rough edges, but the architectural vision is sound and the AWS backing suggests long-term commitment.
