kiroawsbedrock

Kiro: a written spec as the unit of work

Kiro is the coding agent from AWS. It turns a prompt into a written spec first, and then writes the code.

Kiro sorts itself by surface. The same agent runs in the IDE, in the CLI, in the browser, in an iOS app and in Crew, a workspace that runs on your own machine. What sets Kiro apart is the unit of work. A prompt becomes a spec: three files called requirements.md, design.md and tasks.md. The code comes after them.

You pick the model in chat, or you leave it on Auto. The menu has OpenAI, Anthropic and several open-weight models. Whose account serves it has one answer: AWS runs every model through Amazon Bedrock. No option for your own key or a local model is documented. You do not need an AWS account to start. People sign in with GitHub, Google or AWS Builder ID, and companies use IAM Identity Center.

From my course slides (MASE-3, FH Burgenland), read 8 September 2026. Some details on the slide have changed since; the text on this page is current as of September 2026.
From my course slides (MASE-3, FH Burgenland), read 8 September 2026. Some details on the slide have changed since; the text on this page is current as of September 2026.[1]

The family

Kiro IDE[2]

desktop IDE
Desktop editor with chat, specs and hooks, built on Code OSS. It imports VS Code settings and Open VSX plugins.

Kiro CLI[3]

CLI
The agent in the terminal, with a headless mode for CI. It reads the same .kiro folder as the IDE.

Kiro Web[4]

web
Agent in the browser that works across GitHub and GitLab repositories in cloud sandboxes and opens pull requests. Paid plans only.

Kiro for iOS[5]

iOS
Start and steer cloud sessions and review pull requests from your phone. Early access through TestFlight.

Kiro Crew[6]

desktop, web dashboard, CLI, chat
Open-source agent workspace on your own hardware, with memory and scheduled tasks. You reach it from a desktop app, the CLI or chat tools such as Slack and Teams.

Kiro in JetBrains IDEs[7]

JetBrains IDE
Kiro works inside IntelliJ IDEA, PyCharm and other JetBrains IDEs through the Agent Client Protocol (ACP).

Plans

Plan names as of September 2026. How you sign in matters as much as the plan: it decides whether your content may be used for service improvement.

PlanForWhat changes
Kiro Free[8]New users. No credit card needed.50 credits a month. Claude Sonnet 4.5 and open-weight models only, with rate limits. Not available for Enterprise or GovCloud.
Kiro Pro[8]Individual developers.1,000 credits a month, premium models including Auto, Claude Sonnet 5 and Claude Opus 5, add-on credits, and Kiro Web.
Kiro Pro+[8]Active individual developers.2,000 credits a month. Otherwise the same as Pro.
Kiro Pro Max[8]Power users.5,000 credits a month. Otherwise the same as Pro.
Kiro Power[8]Intensive users.10,000 credits a month. Otherwise the same as Pro.
Kiro for enterprise[8]Organisations that sign in through IAM Identity Center or their own identity provider.Central billing through AWS, SSO, usage analytics and admin controls. Content is not used for service improvement.

Which model runs

The user picks a model for chat, or leaves it on Auto. Enterprise admins can limit the list and set a default. Model choice applies to chat only. As of September 2026:

Auto[9]Default. Routes each task to a model and balances quality and cost.
GPT-5.6 Sol, Terra, Luna[9]OpenAI, three tiers. Served from the US for every profile region.
Claude Opus 5[9]Also Opus 4.8, 4.7, 4.6 and 4.5. Paid plans.
Claude Sonnet 5[9]Also Sonnet 4.6, 4.5 and 4.0. Sonnet 4.5 is in the Free plan.
Claude Haiku 4.5[9]Smallest Claude model on the list.
Claude Fable 5.1[9]Limited preview for Enterprise. Admins switch it on. Runs only in US East (N. Virginia).
DeepSeek 3.2, MiniMax M2.5 and M2.1, GLM-5, Qwen3 Coder Next[9]Open-weight models. Some are in the Free plan.

What you can configure

Specs[10]requirements.md holds user stories, acceptance criteria or a bug analysis. design.md holds the architecture. tasks.md holds trackable tasks. Independent tasks run at the same time, in waves.
Steering[11]Markdown files with your project's conventions and standards. Kiro reads them in every session, on every surface.
Hooks[12]Run a shell command or an agent prompt when the agent changes a file, uses a tool or finishes a task. A hook can also block a tool.
MCP[13]Connect external tools and data through MCP servers. Enterprise admins can switch MCP off or allow only listed servers.
Powers[14]Packages of tools, workflows and know-how for one technology. Kiro loads them only when they are relevant.
Skills[15]Reusable instruction packages in the open Agent Skills format. Skills from other compatible tools work too.
Custom agents[16]A configuration file sets which tools an agent may use, what it may do without asking, and which context it loads.
Permissions, checkpoints, .kiroignore[17]Control what the agent can reach, rewind or fork a conversation, and keep secrets and sensitive files away from the agent.
Sign-in[18]GitHub, Google, AWS Builder ID, IAM Identity Center or an external identity provider, on IDE, CLI, web and mobile. API keys for the CLI in CI on paid plans.
Enterprise governance[19]Admins decide on models, MCP servers, API keys, web tools and cloud sessions in the Kiro console.

What happens to your data

Free and individual subscribers (GitHub, Google or AWS Builder ID)[20]Content may be used for service improvement, including model training, by default. You can opt out in Settings under Telemetry and Content.
Enterprise (IAM Identity Center or external identity provider)[20]Content is not used for service improvement, and telemetry is off. Prompt logs, if an admin enables them, go to an S3 bucket in your own AWS account.
Amazon Q Developer Pro subscribers[21]If you sign in to Kiro with your Q Developer Pro credentials, Kiro does not use your content for service improvement.
Retention for abuse detection[20]Free Tier inputs may be kept up to 60 days. Flagged OpenAI GPT traffic and all Claude Fable 5.1 traffic may be kept up to 30 days.

Europe: GDPR, data residency and the AI Act

DPA / GDPR terms[22]Kiro is licensed as AWS Content under the AWS Customer Agreement. AWS says its GDPR DPA with the EU standard contractual clauses is part of the AWS Service Terms and applies automatically.
Data residency in the EU[9]Enterprise profiles can sit in Europe (Frankfurt). Content is then stored there, and inference stays in EU Regions. Exceptions: GPT-5.6 runs in the US, Claude Fable 5.1 only in US East (N. Virginia), and models marked experimental can run in any commercial AWS Region.
Free and individual plans[20]Content is stored in US East (N. Virginia), and models are served from the US. The EU option is for Enterprise only.
AI Act[23]Amazon, Anthropic and OpenAI have signed the EU Code of Practice for general-purpose AI models. The makers of DeepSeek, MiniMax, GLM and Qwen are not on the list. Kiro's own pages say nothing about the AI Act.

What the AI Act and the GDPR ask of you as a user →

Hybrid: the same models in your own account

Kiro has no route to run its models in your own cloud account: AWS runs the inference. What you can bring from your AWS account is your identity, your logs, your keys, your network and the Region.

AWS runs all models[9]Kiro is powered by Amazon Bedrock. No option for your own API key, your own account or a local model is documented.
Frankfurt Enterprise profile[9]Content stored in Europe (Frankfurt), inference in EU Regions, with the model exceptions above: GPT-5.6, Claude Fable 5.1 and experimental models.
Your identity, logs, keys and network[24]IAM Identity Center (also with Okta or Entra ID), prompt logs in your own S3 bucket, customer managed KMS keys and private access through AWS PrivateLink.
Crew on your own hardware[21]Crew runs locally, but it works through the Kiro CLI, so model calls still go through your Kiro subscription.
AWS GovCloud (US)[25]For regulated US workloads, with enterprise sign-in only. Kiro IDE and CLI are HIPAA eligible and in scope of AWS's ISO/IEC 27001 certification.
The same finished product, in three configurations: consumer, business, or on a cloud account you already have.
The same finished product, in three configurations: consumer, business, or on a cloud account you already have.[26]
Recently changed
  • Amazon Q Developer IDE plugins and paid subscriptions reach end of support. AWS points IDE users to Kiro. Amazon Q in the AWS console is not affected (30 Apr 2027).
  • New signups for Amazon Q Developer closed, including the free tier with Builder ID (15 May 2026).

What would help you next?

Learn to decide

The course “From Managed to Hybrid” shows where AI should run: what each layer costs, what you operate and where your data has to stay. In preparation, with a free live session first.

Linda MohamedLinda Mohamed
AI and cloud consultant, AWS Community Hero, Vienna

Roll it out with me

Tell me which tools your teams use or want. In 30 minutes we look at plans, settings, data rules and whether AWS funding can cover part of a pilot. How AWS funding works →

You already know what you need

Book a workshop directly:

  • Strands Agents Workshop →
    One day for developers: build a working agent system with custom tools, MCP and agent-to-agent communication.
  • AI Enablement Workshop →
    One workshop for one role, up to 12 people: how your team uses these tools well, and what stays inside the rules.
ai-solutions.wiki

Read more on ai-solutions.wiki

My open knowledge base, with longer comparisons and sources.

The technology underneath: AI technologies explained →

Sources

  1. Linda Mohamed, course slides “The projects”, appendix layer 0, MASE-3, FH Burgenland, September 2026
  2. Kiro IDE, kiro.dev. https://kiro.dev/docs/ide/
  3. Kiro CLI, kiro.dev. https://kiro.dev/docs/cli/
  4. Kiro Web, kiro.dev. https://kiro.dev/docs/web/
  5. Kiro for iOS, kiro.dev. https://kiro.dev/docs/mobile/
  6. Kiro Crew, kiro.dev. https://kiro.dev/docs/crew/
  7. Kiro in JetBrains IDEs, kiro.dev. https://kiro.dev/docs/upgrade-guides/migrating-from-q-developer/
  8. Kiro Free, kiro.dev. https://kiro.dev/pricing/
  9. Auto, kiro.dev. https://kiro.dev/docs/models/
  10. Specs, kiro.dev. https://kiro.dev/docs/specs/
  11. Steering, kiro.dev. https://kiro.dev/docs/steering/
  12. Hooks, kiro.dev. https://kiro.dev/docs/hooks/
  13. MCP, kiro.dev. https://kiro.dev/docs/mcp/
  14. Powers, kiro.dev. https://kiro.dev/docs/powers/
  15. Skills, kiro.dev. https://kiro.dev/docs/skills/
  16. Custom agents, kiro.dev. https://kiro.dev/docs/custom-agents/
  17. Permissions, checkpoints, .kiroignore, kiro.dev. https://kiro.dev/docs/
  18. Sign-in, kiro.dev. https://kiro.dev/docs/getting-started/authentication/
  19. Enterprise governance, kiro.dev. https://kiro.dev/docs/enterprise/governance/
  20. Free and individual subscribers (GitHub, Google or AWS Builder ID), kiro.dev. https://kiro.dev/docs/privacy-and-security/data-protection/
  21. Amazon Q Developer Pro subscribers, kiro.dev. https://kiro.dev/faq/
  22. DPA / GDPR terms, aws.amazon.com. https://aws.amazon.com/compliance/gdpr-center/
  23. AI Act, digital-strategy.ec.europa.eu. https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai
  24. Your identity, logs, keys and network, kiro.dev. https://kiro.dev/docs/privacy-and-security/vpc-endpoints/
  25. AWS GovCloud (US), kiro.dev. https://kiro.dev/docs/privacy-and-security/compliance-validation/
  26. Linda Mohamed, course slides “Start here”, MASE-3, FH Burgenland, September 2026
  27. Kiro, ai-solutions.wiki. https://ai-solutions.wiki/glossary/kiro/
  28. Claude Code vs Cursor vs Codex, ai-solutions.wiki. https://ai-solutions.wiki/comparisons/claude-code-vs-cursor-vs-codex/
  29. AWS: Amazon Q Developer end-of-support announcement. https://aws.amazon.com/blogs/devops/amazon-q-developer-end-of-support-announcement/
  30. AWS docs: Amazon Q Developer IDE plugins end of support. https://docs.aws.amazon.com/amazonq/latest/qdeveloper-ug/q-developer-ide-end-of-support.html
  31. Kiro license (AWS Content). https://kiro.dev/license/

Checked against the vendors' own pages in September 2026. Plans, models and features change often: check the linked pages before you decide.