Kiro CLI Setup

Version: 0.10

Multi-agent AWS development environment for the Kiro CLI — master orchestrator, 12 specialist subagents, 16 always-loaded steering docs, 24 skills, and a curated MCP server stack centered on the AWS Agent Toolkit.

💡 Tips for AI Agents working on this repo

• NEVER push directly to main without running ./validate.sh first. It validates every agent JSON, JSON syntax, bash syntax, and the gitignore privacy guard. The script prints ✓ All checks passed. Safe to push. when green. Run after ANY change to agents/, prompts/, skills/, steering/, settings/, or import.sh.
• Prefer the configured MCP servers over CLI commands (gh, aws, curl, etc.) for the same operation. See steering/mcp-server-preference.md for the full mapping table.
• Don't recommend CI/CD pipelines or git hooks — both are explicitly banned in steering/post-task-recommendations.md. Validation belongs in ./validate.sh and deploy.sh, not automation hooks.
• When generating recommendations, follow the user/AI Agent split in steering/post-task-recommendations.md.

See Releases for the latest version notes.

Prerequisites

• Kiro CLI installed (latest tested with 2.4.1)
• Python 3.14+ with uv and uvx
• Node.js 20+ with npx
• AWS CLI v2 configured with named profiles
• Git

Local Tooling Required by MCP Servers

Several MCP servers shell out to local tools. If you're using an AI coding agent to set this config up automatically, the agent must install everything below before any MCP server will work. The included import.sh handles all of this on macOS.

Rule of thumb for AI agents setting this up autonomously: if import.sh is available, run it — it installs everything above interactively. If you must script it from scratch, install Homebrew first, then uv, then node, then awscli, then awsdac, then the formatter tools — in that order.

One-shot install (macOS, AI-agent friendly)

# Homebrew
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

# uv (Python package runner)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Everything else via Homebrew
brew install node awscli awsdac graphviz ruff shfmt git-delta
npm install -g prettier

# Then run the importer (handles MCP keys + Google Workspace OAuth + agent installs)
./import.sh

Clone the repo into your Kiro config directory:

git clone https://github.com/seandkendall/kiro-config.git ~/.kiro
cd ~/.kiro

Start Kiro CLI:

kiro-cli chat -a -r

Then paste this prompt:

Please review my Kiro/Kiro CLI config in this directory including all agents,
subagents, configurations, steering docs, skills, prompts, and all MCP servers
we have configured in the various files.

First, research all changes AWS has made to Kiro CLI in this changelog:
https://kiro.dev/changelog/cli/
and the IDE in this changelog:
https://kiro.dev/changelog/ide/
available models in this changelog:
https://kiro.dev/changelog/models/
and other general changes in this changelog:
https://kiro.dev/changelog/general/

Next, research the AWS AIDLC GitHub page to see if there is anything valuable
from this project to port into our Kiro configuration:
https://github.com/awslabs/aidlc-workflows/tree/main

Next, AWS has released the Agent Toolkit for AWS which you can find here:
https://aws.amazon.com/products/developer-tools/agent-toolkit-for-aws/
and the GitHub for this here:
https://github.com/aws/agent-toolkit-for-aws

Deep research these projects to see what we should be pulling into our Kiro
configuration. Also make sure you are looking at the changelogs to see what
is new and how that can make our Kiro environment better.

For any MCP server that requires an API key, check to see if I have a key.
If I do not have a key on my machine, then ask me for it, and if I decline,
then simply remove that MCP server from the configuration.

# Clone directly into your Kiro config directory
git clone https://github.com/seandkendall/kiro-config.git ~/.kiro

# Or clone elsewhere and copy over
git clone https://github.com/seandkendall/kiro-config.git
cp -r kiro-config/{agents,steering,skills,prompts,settings} ~/.kiro/

What's Included

Agents (17)

Steering Docs (16)

Rules and standards automatically loaded into every session: accessibility, API design, AWS/CDK patterns, AWS Agent Toolkit usage, development workflow, error handling, performance, Python standards, security policies, and more.

Personal steering — say "always …" / "from now on …" / "I prefer …" or repeat a preference 2+ times in a session, and the agent will offer to save it as a personal-<topic>.md steering doc on your machine. These files are gitignored and ALWAYS win over base rules. No extra tooling required — they just work via Kiro's existing steering loader. See steering/personal-rules-protocol.md and skills/personal-rules-management.md.

Skills (24)

MCP Servers

Agents configure their own MCP servers. Key servers used across agents:

Environment Variables

The setup uses 3 optional API keys. If a key is missing when you run ./import.sh, the installer will prompt you for it and disable the related MCP server if you skip.

Add them to ~/.zshrc:

export GITHUB_PERSONAL_ACCESS_TOKEN="ghp_your_token_here"
export TWENTY_FIRST_API_KEY="your_key_here"
export FIGMA_API_KEY="your_key_here"

Then source ~/.zshrc (or open a new terminal) before starting Kiro CLI.

Configuration

The default model and settings are in settings/cli.json. Key settings:

• chat.defaultAgent: master (the orchestrator)
• chat.defaultModel: claude-opus-4.8
• chat.enableSubagent: true
• chat.enableThinking: true
• chat.enableTodoList: true
• toolSearch.enabled: true

Key Conventions

• CDK in Python only — never TypeScript for infrastructure
• TypeScript for React frontends only
• deploy.sh is the only deployment method — no CI/CD pipelines, no git hooks
• Kiro Specs before code — requirements.md → design.md → tasks.md
• 10+ recommendations after every completed task
• cdk-nag for security validation on all stacks
• Lambda Powertools (Logger, Tracer, Metrics) on all Lambda functions
• MCP-over-CLI — github MCP for github.com operations (never gh); aws-mcp-server for AWS (never bare aws CLI)

Troubleshooting

Common install issues:

For deeper Kiro CLI issues (settings not loading, side channels in deploy.sh, agent validation errors), see steering/kiro-cli-troubleshooting.md.

License

Apache License 2.0 — see LICENSE. Shared publicly for reference and reuse. See CONTRIBUTING.md for the pre-push validation flow and contribution guidelines.