Kiro

Kiro is an agentic AI development environment from a small, focused team inside Amazon Web Services (AWS).

Designed to help engineers write, debug, and understand code more efficiently, Kiro is more than a coding copilot — it’s an agentic development environment that reimagines how developers collaborate with AI.

Powered by Anthropic’s Claude models and hosted securely within AWS infrastructure, Kiro turns developer prompts into structured specs, designs, and tasks — enabling AI agents to contribute real engineering work. Through features like agent hooks, developers can delegate complex tasks like generating tests, optimizing performance, or updating documentation — all in the background.

Kiro structures AI coding around specs, designs, and tasks, so agents can do real engineering work with you—not just chat.

Kiro’s workflow centers on spec-driven development and agent hooks:

• Specs turn a prompt into concrete requirements, architecture, and a task list.
• Agent hooks let you automate steps like tests, docs, or perf tweaks on events (e.g., file save). You stay in control with diffs and approvals while agents execute the plan.

Under the hood, Kiro runs on Claude Sonnet models (3.7 and 4), with more options planned.

Kiro Pricing

Kiro bills by request type to match how you build:

• Vibes = chat/assist interactions.
• Specs = executing tasks/subtasks from your spec workflow.

Monthly plans

• Free — $0, 50 Vibes, 0 Specs
• Pro — $20, 225 Vibes, 125 Specs
• Pro+ — $40, 450 Vibes, 250 Specs
• Power — $200, 2,250 Vibes, 1,250 Specs

Overages (paid plans only): $0.04 per Vibe, $0.20 per Spec.
Welcome bonus: +100 Vibes +100 Specs for the first 14 days on new accounts.
Note: Taxes may apply.\

Notes:

Currently in limited release, Kiro has seen unprecedented demand, prompting a waitlist to ensure a smooth onboarding experience. Existing users can continue using it, and new users are being added in batches as the team rapidly scales and improves the platform.

Kiro is Amazon’s Claude-powered answer to tools like GitHub Copilot and Code Assist — but with deeper structure, more autonomy, and a relentless focus on building the future of software development with AI.