feat/new-feature
Product engineering
New features, specified first
A product brief becomes requirements, design and tasks. Agents build and test each step while you review.
$ kiro --spec-driven
Kiro is the agentic IDE and CLI from AWS. Before it writes a line of code, it drafts requirements, a design and a task list that your team can read, question and fix. We help engineering teams roll it out properly: specs, steering, hooks, MCP, and the review habits that make it stick.
Explorer
.kiro
specs/refunds
requirements.md
design.md
tasks.md
steering
hooks
src
tests
agentworking on task 3 of 5
hookon save: tests/refunds.spec.ts updated ✓
review2 files changed, waiting for you
Talk to an engineer
Tell us a little about your team and your codebase. One of our engineers will get back to you within one business day. If Kiro is not a good fit for what you are doing, we will tell you.
README.md
Most AI coding tools go straight from a prompt to a pile of code. Fine for a demo. Less fine three weeks later, when someone asks why the refund service retries five times and nobody remembers deciding that.
Kiro works the other way round. You describe what you want and it writes a spec first: requirements with acceptance criteria, a technical design, and an ordered list of tasks. You review those, change what is wrong, and only then does the agent work through the tasks, writing code and tests as it goes.
The specs are plain Markdown files in your repo, so the reasoning sits next to the code and goes through review like everything else. Kiro is built on Code OSS, so it feels like VS Code: your settings, themes, keybindings and Open VSX extensions come with you. There is a CLI as well, for people who live in the terminal.
workflow.diff
.kiro/specs
This is the loop Kiro runs on. Each stage is a file you can open, edit and commit.
Your prompt becomes user stories with acceptance criteria, written in EARS notation. Vague asks turn into statements you can actually test.
WHEN a customer requests a refund within 30 days THE SYSTEM SHALL create a refund record and notify payments
Kiro reads your codebase and proposes the architecture: components, data models, interfaces and error handling, before anyone starts typing.
## Components - RefundController - RefundService - PaymentsQueue (SQS) ## Errors: retry x3, then DLQ
The design turns into an ordered checklist, each task tied back to a requirement. Run them one at a time and watch what happens.
[x] 1. Refund model + migration [x] 2. RefundService.create() [ ] 3. SQS notification [ ] 4. Integration tests
The agent works the list, writes code and tests, and shows you the diffs. Supervised mode asks before each change. Autopilot just gets on with it.
+ src/refunds/service.ts + tests/refunds.spec.ts ~ docs/refunds.md (hook) ready for review
kiro --help
Kiro mixes structured planning with agents that do the work. Here is the feature list, minus the marketing.
Ask us how it fits your stackUSAGE kiro [feature] --for your-team
engagements/
Buying seats is the easy bit. Getting a team to write good specs and trust the output takes a bit more. Pick the engagement that matches where you are.
Before anyone buys licences, we look at how your teams actually ship: repos, toolchain, review process. Then we pick one or two real projects and run Kiro on them, measuring before and after so the decision rests on evidence rather than gut feel.
We set Kiro up once, properly, for the whole organisation: shared steering files, an approved list of MCP servers, a library of reusable hooks, and clear rules for secrets and secure coding. New teams start from that baseline instead of a blank folder.
Kiro is only as good as the specs people write and the reviews they do. Our hands-on workshops cover both, for developers, tech leads and product owners, and we build them around your own repos wherever we can.
You know that 3,000-line service nobody wants to touch? Our engineers use Kiro to document it first, put tests around it, and then refactor towards cloud-native patterns on AWS, one reviewed step at a time.
AWS is pointing Amazon Q Developer IDE users towards Kiro. If that is you, we review how your developers use Q today, map plans and subscriptions, move workflows and customisations across, and train people so the switch does not stall a sprint.
git branch -a
feat/new-feature
Product engineering
A product brief becomes requirements, design and tasks. Agents build and test each step while you review.
chore/compliance
BFSI
Steering files carry your secure coding and audit rules. Hooks run the checks on every change, so nobody has to remember.
refactor/legacy
Legacy modernisation
Document legacy applications, then refactor them to modern frameworks with tests wrapped around them first.
infra/iac
Cloud & DevOps
Write and review CloudFormation, CDK and Terraform faster, with the spec recording why the architecture looks the way it does.
test/coverage
Quality engineering
Hooks update unit and integration tests whenever code changes, so coverage stops quietly slipping.
main
Startups & ISVs
Ship the quick version without painting yourself into a corner. The specs are already there when engineer number five joins.
about/consultedge
We are an AWS Advanced Tier Partner with our own software engineering practice, and our teams use Kiro on real delivery work. The advice you get on specs, steering and hooks comes from doing it, including finding out what does not work.
$ consultedge rollout kiro --team yours
✓ done. Your team owns it from here, and we stay on call.
FAQ.md
Something missing? Ask us directly. We would rather answer the real question than the tidy one.
Kiro is an agentic IDE and CLI from AWS built around spec-driven development. You describe what you want, Kiro turns it into requirements, a design and a task list, and then AI agents implement, test and document the work. The aim is to get from prototype to production without losing track of why the code looks the way it does.
It means agreeing on what to build before building it. Kiro writes a requirements document with acceptance criteria, a technical design and a list of implementation tasks. Those specs stay in your repository, so every piece of AI-generated code can be traced back to a stated intent.
Agent hooks are automations that run on events such as saving, creating or deleting a file, for example updating tests or documentation. Steering files are Markdown files that tell Kiro about your architecture, conventions and preferred libraries, so its output fits your codebase.
Yes. Kiro is built on Code OSS, so developers can import their VS Code settings, themes and keybindings and use compatible extensions from the Open VSX registry. The real learning curve is writing and reviewing specs, not the editor.
AWS is positioning Kiro as the next step for AI-assisted development, and Amazon Q Developer users are being guided toward it. We can plan and run that move for you, covering plan mapping, workflow migration and training, so it does not disrupt delivery.
Kiro is built by AWS with enterprise controls, and organisations can manage access through their identity provider. As part of our governance setup we also review data-handling settings, which MCP servers the agent can reach, and how secrets are managed. Tell us your compliance requirements and we will be straight with you about what fits.
Kiro has a free tier plus paid individual and team plans based on monthly usage credits. AWS adjusts plans from time to time, so it is worth checking current pricing. We help you pick the right mix for your developers as part of adoption planning.
No. Any developer can download Kiro and start today. Where we help is the organisation-wide part: governance, shared steering and hooks, training, and a pilot that gives you numbers you can act on.