Overview
Kiro is AWS’s agentic development environment, built on a contrarian premise: AI already writes code faster than teams can review it, so the bottleneck is trust rather than speed. Instead of turning a prompt straight into a diff, Kiro negotiates a spec with you first — structured requirements, a system design that surfaces architectural trade-offs, and an ordered task list — and only then dispatches agents to implement it. What is left behind is a written record of why the code looks the way it does, which is precisely what code review and audit normally lack after a vibe-coding session.
It ships in three shapes: a desktop IDE for macOS, Windows, and Linux built on Code OSS, so VS Code settings, themes, and Open VSX extensions import during onboarding; a terminal CLI; and a web interface on paid plans. No AWS account is required — you sign in with GitHub, Google, or an AWS Builder ID. Around the spec core sit steering files that encode house coding standards, agent hooks that fire autonomous agents on events such as a file save, persistent memory across sessions, and Model Context Protocol support for wiring in external tools.
Kiro reached general availability in November 2025 and is AWS’s declared successor to Amazon Q Developer, which has stopped taking new signups and reaches end of support in April 2027. Model access spans open-weight options such as Qwen3 Coder Next, DeepSeek v3.2, and MiniMax alongside Claude Sonnet and Opus tiers, routed by default through an Auto mode that picks a model per task.
Billing is credit-metered rather than flat-rate. That is unusually transparent — you can see what each task costs — but harder to budget: Free grants 50 credits a month, Pro is $20 for 1,000, and the ladder runs to Power at $200 for 10,000, with add-on credits at $0.04 each. Kiro fits teams that need AI output to survive review and audit; it fits solo tinkerers less well.
Key Features
- Spec-driven development: requirements, system design, and sequenced tasks before any code
- Agent hooks that trigger autonomous agents on events such as file saves or commits
- Steering files to encode team coding standards and constraints persistently
- Available as desktop IDE (Code OSS based), terminal CLI, and web interface
- Model choice across open-weight and premium Claude tiers, with an Auto routing mode
- MCP support plus cross-session memory for connecting external tools and context
Pricing
| Plan | Price | For |
|---|---|---|
| Free | $0 | 50 credits/mo, open-weight models + Claude Sonnet, rate limited |
| Pro | $20/user/mo | 1,000 credits, premium models, $0.04 add-on credits |
| Pro+ | $40/user/mo | 2,000 credits, heavier daily agentic use |
| Pro Max | $100/user/mo | 5,000 credits, large codebases and parallel tasks |
| Power | $200/user/mo | 10,000 credits, maximum throughput |
| Enterprise | Contact sales | SSO via IAM Identity Center, usage analytics, consolidated billing |
Comparison
Compared to Cursor, Kiro trades inline iteration speed for documented intent — Cursor is the faster editor for tight feedback loops, while Kiro produces a spec artifact your reviewers can actually read. Against Claude Code, both are agentic and terminal-capable, but Claude Code optimizes for autonomous multi-file execution under a flat subscription, whereas Kiro meters credits and gates work behind an approved plan. Next to Windsurf, Windsurf’s Cascade agent keeps you inside a conventional IDE flow, while Kiro’s hooks and steering files target engineering organizations that need governable, repeatable agent behavior.