Kiro Installation Guide

Kiro is AWS's agentic development platform. Instead of prompting an assistant to write code, you write a spec — requirements, design and tasks — and the agent implements against it.

System Requirements

• macOS, Windows or Linux on a 64-bit processor
• 8 GB RAM or more (16 GB recommended for large repositories)
• 5 GB of free disk space
• A Kiro account with the IDE, CLI or web surface you want to use
• A terminal and git for the CLI and for working with the project

Installation Instructions

The IDE

Kiro is downloaded from the official downloads page, which detects your platform:

# macOS — download the .dmg from https://kiro.dev/downloads/
# and drag Kiro into your Applications folder
 
# Windows — download the .exe installer from https://kiro.dev/downloads/
# and run it
 
# Linux — download the .deb or .rpm from https://kiro.dev/downloads/
sudo apt install ./kiro_amd64.deb
# or
sudo dnf install ./kiro-amd64.rpm

All platform downloads, including the AppImage and the ARM64 builds, are listed on kiro.dev/downloads.

The CLI

The CLI gives you the same agent in a terminal, which is what you want for headless automation and CI/CD:

# Follow the official setup for your platform
# https://kiro.dev/docs/cli/setup/

The CLI is also the fastest way to use Kiro from a container or a remote host, where the IDE is not an option.

Signing In

Kiro needs a Kiro account:

# The CLI walks you through the browser flow
kiro auth login
 
# Check the current state
kiro auth status

The Kiro Workflow

1. Write a Spec

Kiro's core idea is that a spec is a living document with three parts:

• Requirements — what the feature must do, expressed as numbered requirements
• Design — how it will be built, including the architecture and the trade-offs
• Tasks — the implementation steps, each traceable back to a requirement

Kiro can help you draft the first version:

Create a spec for adding two-factor authentication to the login flow. Requirements
first, then the design, then the tasks.

2. Run Requirements Analysis

Before any code is written, Kiro checks the spec against itself and finds the gaps:

# From the CLI
kiro spec analyze-requirements

This surfaces contradictions, ambiguities and missing acceptance criteria while they are still cheap to fix.

3. Implement Task by Task

# Generate the implementation for the next task
kiro task start
 
# Check that the code matches the spec
kiro spec verify

4. Verify with Property-Based Tests

Kiro can turn requirements into executable properties — checks that hold for generated inputs, not just the examples you wrote:

kiro spec correctness

Steering: Project Conventions

Steering files are persistent instructions that shape every request in a project. They live in .kiro/steering/ and are committed to Git:

<!-- .kiro/steering/backend-conventions.md -->
 
- All API handlers live in `src/routes/` and use the `createHandler` factory.
- Every database access goes through the repository layer in `src/db/`.
- Run `npm run lint && npm test` before a task is marked complete.

Keeping the Agent in Control

Troubleshooting

Best Practices

1. Write the spec before the code — the requirements analysis step catches more problems than any amount of prompt tweaking.
2. Keep steering files in Git so the whole team gets the same behaviour.
3. One task at a time — it keeps the diff reviewable and the agent focused.
4. Use property-based correctness checks for the requirements that matter.
5. Set permissions deliberately — start with the agent restricted to the project directory.
6. Commit the .kiro/ directory except for anything machine-specific.
7. Use cloud sessions for long refactors you can hand off and come back to.

Next Steps

• Kiro Documentation
• Kiro Downloads
• Kiro on GitHub
• Cursor
• Windsurf / Devin Desktop
• Claude Code