Kiro
Amazon Web Services · Kiro: AWS's Spec-Driven Agentic Coding IDE, CLI, and Cloud Agent
Kiro is an agentic development environment from AWS that turns a prompt into a spec (requirements, design, and a task list) and then implements it with agents. It runs as a desktop IDE, a CLI, and a web app with cloud sessions, and suits developers and teams who want planned, reviewable changes rather than free-form prompting.
Best for
Developers and teams who want planned, documented changes (requirements, design, and tasks) before code is written, especially for multi-step features and teams already using AWS identity
Not ideal for
Developers who mostly want quick inline edits or autocomplete, teams that need an open-source or self-hosted agent, or free-tier users who need cloud sessions
Who it's for
Software developers and engineering teams who want spec-first, reviewable agentic coding across an IDE, CLI, and cloud sessions
Capabilities
- Spec-driven development: generates requirements.md (or bugfix.md), design.md, and tasks.md before implementation, with feature specs and bugfix specs
- Parallel task execution: independent spec tasks run concurrently in waves while dependent tasks wait, with progress tracked in tasks.md
- Agent hooks that run shell commands or agent prompts on events such as file save, create, or delete, prompt submit, agent stop, and before or after tool use (PreToolUse can block)
- Steering files in `.kiro/steering/` with always, fileMatch, manual, and auto inclusion modes, plus AGENTS.md support
- Property-based testing to check correctness across many inputs (IDE only)
- Kiro Web and cloud sessions that run in a remote sandbox, continue while you are disconnected, coordinate changes across GitHub and GitLab repositories, and can be resumed from browser, IDE, CLI, or mobile
- Kiro CLI with headless mode for CI/CD, custom agents, voice mode, and MCP support
- Model choice including Auto, Claude models (Opus 5, Sonnet 5, and earlier), OpenAI GPT-5.6 Sol, Terra, and Luna, and open-weight models such as Qwen3 Coder Next, DeepSeek 3.2, GLM-5, and MiniMax
- Shared `.kiro/` configuration across IDE, CLI, web, and mobile, with cloud configuration to reuse a local setup in cloud sessions
Limitations
- The free tier is limited to 50 credits a month, Claude Sonnet 4.5 and open-weight models, and does not include Kiro Web or cloud sessions
- Credit consumption varies with prompt complexity and model (spec tasks typically cost more than one credit), and unused monthly credits do not roll over (purchased add-on credits expire after 12 months)
- Kiro Web and cloud sessions are offered in one AWS Region, US East (N. Virginia)
- Subscriptions cannot be used through third-party harnesses (such as OpenClaw) that route requests outside Kiro's own IDE, CLI, web, Crew, and ACP interfaces
- The iOS app is in early access, and no Android app is documented
- Kiro itself is not open source; only Kiro Crew, a separate persistent workspace tool, is released under Apache-2.0
- Kiro's docs state that supervised mode is a code review workflow, not a security control; limiting what the agent can touch relies on protected paths, trusted commands, workspace isolation, and credential scoping
- The IDE installs extensions from the Open VSX registry, so some VS Code Marketplace-exclusive extensions are not available
- The spec workflow adds requirements and design steps that are more process than small, one-off edits need
Use cases
- Turning a feature request into requirements, a design doc, and a task list, then letting Kiro implement the tasks while you review each step
- Diagnosing a bug with a bugfix spec that captures root-cause analysis before changing code
- Adding hooks that run linters, formatters, or type checks automatically after the agent edits files
- Starting a cloud session for a change that spans several GitHub and GitLab repositories and checking on it later from the IDE or phone
- Running Kiro CLI in headless mode inside a CI pipeline
Our take
Kiro's value is structure. Specs, steering files, and hooks turn a coding agent into a repeatable process that leaves a paper trail, which helps on multi-step features and on teams that review plans before code. Cloud sessions you can resume from the IDE, terminal, or phone make long tasks easier to hand off. The costs are overhead and predictability: specs are more process than small edits need, credits burn at different rates depending on model and task, and cloud sessions are paid-only and run in a single AWS region. It suits teams that want planned agent work more than developers who mostly want fast inline help.
Who should use it
Teams that want agents to plan before coding, developers building multi-step features who want reviewable requirements and design docs, organizations standardizing conventions through steering files and hooks, and AWS-centric teams that sign in with IAM Identity Center.
Who should skip it
Developers who mainly want autocomplete or quick edits, teams that require an open-source or self-hosted agent, organizations that need cloud sessions outside US East, and budget-constrained users who need predictable flat usage.
Strengths
- Spec-first workflow leaves requirements, design, and task files you can review and keep in the repo
- Hooks and steering files give teams repeatable automation and conventions
- One configuration across IDE, CLI, web, and mobile, with cloud sessions you can resume anywhere
- Free tier and published per-credit overage pricing
- Enterprise sign-in through AWS IAM Identity Center or an external identity provider
Weaknesses
- Credit consumption varies by model and task, so monthly usage is harder to predict
- Cloud sessions are paid-only and offered only in US East (N. Virginia)
- Spec planning adds overhead for small changes
- Not open source and no self-hosted option documented
Kiro pricing
Kiro Free
$0
Billed monthly
- 50 credits per month
- Claude Sonnet 4.5 and open-weight models
Kiro Pro
$20/user/month
Billed monthly
- 1,000 credits
- Premium models
- Add-on credits at $0.04 each
Kiro Pro+
$40/user/month
Billed monthly
- 2,000 credits
- Premium models
- Add-on credits at $0.04 each
Kiro Pro Max
$100/user/month
Billed monthly
- 5,000 credits
- Premium models
- Add-on credits at $0.04 each
Kiro Power
$200/user/month
Billed monthly
- 10,000 credits
- Premium models
- Add-on credits at $0.04 each
Enterprise
Contact sales
- Centralized billing, SSO, usage analytics, and enterprise security controls
Free tier limits: 50 credits per month; Claude Sonnet 4.5 and open-weight models only; Kiro Web and cloud sessions not included
Note: A credit is a unit of work: simple prompts can use less than one credit, while spec tasks typically use more. Premium models consume credits faster than Auto. Usage resets each billing month and unused plan credits do not roll over; add-on credits (social login or Builder ID) roll over and expire 12 months after purchase, while Enterprise admins can enable overages billed at $0.04 per credit. Cloud session compute draws from the same credits. AWS GovCloud pricing is about 20% higher and has no free tier.
Technical specs
Available models
Where Kiro excels
Building a multi-step feature with a reviewable plan
Kiro produces requirements, a design, and a task list that a lead can review before any code changes, then implements the tasks and tracks progress in tasks.md.
Enforcing team conventions automatically
Steering files load project standards into agent interactions, and hooks can run formatters, linters, or type checks after every agent file change.
Handing off a cross-repo change to the cloud
A cloud session can coordinate changes across GitHub and GitLab repositories, keep running after you disconnect, and be resumed later from the browser, IDE, or terminal.
Kiro vs. competitors
Kiro vs. Cursor
Cursor is an AI code editor where the agent works mostly from chat and prompts, with cloud agents for background work. Kiro is also an agentic IDE, but its core workflow is writing requirements, a design, and tasks before code, plus event-driven hooks.
Kiro vs. Google Antigravity
Google Antigravity is an agent-first platform centered on running many coding agents in parallel from a desktop command center. Kiro emphasizes structured specs and hooks, with cloud sessions you can resume across IDE, CLI, web, and mobile.
Kiro vs. Claude Code
Claude Code is Anthropic's coding agent for Claude models in the terminal, IDEs, desktop, and web. Kiro offers Claude models among others and puts a spec-driven planning layer and hooks around the agent.
Kiro vs. OpenAI Codex
OpenAI Codex spans an open-source CLI, IDE extensions, and parallel cloud tasks tied to ChatGPT plans. Kiro uses credit-based plans, offers Claude, OpenAI GPT-5.6, and open-weight models, and structures work as specs before implementation.
Frequently asked questions
What is Kiro?
Kiro is an agentic development environment built and operated by AWS. It turns prompts into specs (requirements, design, and tasks) and implements them with agents across a desktop IDE, a CLI, a web app with cloud sessions, and an iOS app in early access.
How much does Kiro cost?
Kiro Free is $0 with 50 credits a month. Paid plans are Pro at $20/user/month (1,000 credits), Pro+ at $40 (2,000), Pro Max at $100 (5,000), and Power at $200 (10,000), with add-on credits at $0.04 each. Enterprise pricing is through sales.
What is spec-driven development in Kiro?
Kiro writes a requirements.md (or bugfix.md for bugs), a design.md, and a tasks.md before implementing. It then works through the tasks, running independent tasks in parallel and marking progress in tasks.md, so the plan and its execution stay reviewable.
Which models does Kiro support?
By default, Auto routes each prompt across a mix of models to balance quality, latency, and cost. Free users get Claude Sonnet 4.5 and open-weight models such as Qwen3 Coder Next, DeepSeek 3.2, and MiniMax M2.1. Paid plans add premium models including Claude Opus 5, Claude Sonnet 5, and OpenAI's GPT-5.6 Sol, Terra, and Luna. Models consume credits at different rates relative to Auto.
Is Kiro open source?
No. Kiro's IDE, CLI, and web app are proprietary AWS products. Kiro Crew, a separate persistent development workspace, is open source under Apache-2.0.
Does Kiro replace Amazon Q Developer CLI?
Yes. Kiro's upgrade guide describes Kiro CLI as the next update of the Amazon Q Developer CLI: Q CLI installs with auto-update enabled were moved to Kiro CLI in November 2025, and existing Q Developer CLI workflows, subscriptions, and sign-in keep working. Kiro CLI adds headless mode for CI, custom agents, and MCP support.
Integrations & fit
Alternatives to consider
Anysphere
VS Code developers who want multi-file AI editing, autonomous cloud agents, and model flexibility — not a plugin layered on top of an existing editor
Freemium
Developers who want to supervise several autonomous coding agents in parallel from a dedicated desktop app, with a free tier and access to both Gemini and Claude models
FreemiumAnthropic
Developers and teams who want deep codebase reasoning, multi-file changes, and persistent project instructions in a terminal-friendly or IDE-native workflow
Paid
OpenAI
Developers and teams already paying for ChatGPT Plus, Pro, Business, or Enterprise who want one coding agent across the terminal, IDE, and parallel cloud tasks
FreemiumAbout Kiro
Kiro's central idea is spec-driven development. Before writing code, the agent produces three files: requirements.md (user stories with acceptance criteria, or bugfix.md for bug analysis), design.md (architecture and sequence diagrams), and tasks.md (trackable implementation tasks). It then executes independent tasks in parallel waves and marks each one in progress or complete. Around that core are steering files in `.kiro/steering/` that encode project conventions (AGENTS.md is also supported), agent hooks that run shell commands or agent prompts on events such as file saves, prompt submission, or tool use, and MCP servers for external tools. In the IDE, Kiro can also check correctness with property-based tests. The same agent harness and `.kiro/` configuration power every surface: a desktop IDE for macOS, Windows, and Linux built on VS Code's open-source foundation (it imports VS Code settings and Open VSX extensions), a terminal CLI with headless mode for CI (Kiro's docs call it the next update of the Amazon Q Developer CLI), Kiro Web for multi-repo tasks that open pull requests, and an iOS app in early access. Since September 3, 2026, Kiro Web and cloud sessions have been generally available on paid plans. A session runs in a remote sandbox in the US East (N. Virginia) AWS Region, keeps working after you disconnect, can change repositories across GitHub and GitLab in one session, and can be resumed from the browser, IDE, terminal, or phone. Pricing is credit-based. The free tier includes 50 credits a month with Claude Sonnet 4.5 and open-weight models, and paid plans run from $20 to $200 per user per month with 1,000 to 10,000 credits, premium models such as Claude Opus 5 and OpenAI's GPT-5.6, and add-on credits at $0.04 each. The main tradeoff is structure: specs add planning steps that pay off on larger features but are more than a quick edit needs, and credit usage varies with model choice and task complexity.
Updates from Kiro
Kiro CLI 2.23.0 added reasoning controls and cloud sessions to the terminal. CLI 2.24.0 followed on September 23 with session-wide tool approval and environment isolation.
AWS made Kiro Web, cloud sessions, and cloud configuration generally available on Pro, Pro+, Pro Max, and Power plans. Cloud sessions run the agent in a remote sandbox in US East (N. Virginia), keep working while you are disconnected, and use the same credits as the rest of Kiro.
Are you the founder? Claim this listing →