
I Tried Out Kiro Crew on an Amazon Linux 2023 EC2 Instance
This page has been translated by machine translation. View original
Introduction
Kiro Crew was announced on the official blog on August 4, 2026 (local time).
For installation on macOS and Slack integration, please see Iwasa's article.
This article introduces installing Kiro CLI and Kiro Crew on an Amazon Linux 2023 EC2 instance, starting the Gateway, and using the Dashboard via an SSH tunnel.
Verification
Verification Environment
The verification was performed in the following environment.
| Item | Value |
|---|---|
| EC2 instance type | t3a.medium |
| Architecture | x86_64 |
| OS | Amazon Linux 2023 |
| Region | ap-northeast-1 |
| kiro-cli | 2.16.1 |
| kirocrew | 0.1.2 |
| Build method | CloudFormation (IAM stack + EC2 stack) |
The full CloudFormation templates are included in the reproduction information at the end. They are for x86_64 t3a.medium.
Setting Up Kiro Crew on an Amazon Linux 2023 EC2 Instance
Deploy the IAM stack and then the EC2 stack with CloudFormation.
export AWS_REGION=ap-northeast-1
export KEY_NAME=<EC2_KEY_PAIR_NAME>
export IAM_STACK_NAME=kirocrew-iam
export EC2_STACK_NAME=kirocrew
aws cloudformation deploy \
--region "$AWS_REGION" \
--template-file kirocrew-iam.yaml \
--stack-name "$IAM_STACK_NAME" \
--capabilities CAPABILITY_NAMED_IAM
aws cloudformation deploy \
--region "$AWS_REGION" \
--template-file kirocrew-ec2.yaml \
--stack-name "$EC2_STACK_NAME" \
--parameter-overrides \
KeyName="$KEY_NAME" \
IamStackName="$IAM_STACK_NAME"
After the installation of kiro-cli and kirocrew finishes via the EC2 UserData, log in over SSH.
Log in to Kiro CLI
kiro-cli login
Enter your organization's Start URL and region, then approve the device code in the browser to log in. Note that the Start URL, device code, and authentication URL are not shown here.
The procedure for signing in with IAM Identity Center is introduced in the article below.
Initial Setup of Kiro Crew
kirocrew setup
This time I used the default workspace. I skipped Slack integration and launching on AWS, and specified Asia/Tokyo for the timezone. The flow of the interaction is as follows.
Interaction log of kirocrew setup
Kiro Crew Setup 👻
── Workspace Directory ──
Default: /home/ec2-user/workplace/kirocrew-workspace
Workspace path [/home/ec2-user/workplace/kirocrew-workspace]:
✅ Workspace: /home/ec2-user/workplace/kirocrew-workspace
Installing agent config...
✅ Agent installed: /home/ec2-user/.kiro/agents/kirocrew.json
── Slack Credentials ──
Configure Slack tokens? [Y/n]: n
⏭ Skipped. Slack integration will be disabled.
── Slash Command ──
Slash command name [kirocrew]:
✅ Slash command: /kirocrew
── Timezone ──
IANA timezone (e.g. America/Los_Angeles): Asia/Tokyo
✅ Timezone saved: Asia/Tokyo
── Run on AWS (optional) ──
Launch KiroCrew on AWS now? [y/N]: n
⏭ Skipped. Launch later: kirocrew cloud launch
👻 Done! Try: kirocrew doctor && kirocrew gateway
I completed the Kiro CLI login and the initial setup of Kiro Crew.
Starting the Gateway and Connecting via an SSH Tunnel
Start the Gateway on the EC2 instance.
kirocrew gateway
Startup log of kirocrew gateway
👻 Kiro Crew gateway starting…
⚠️ Do not enter sensitive, secret, or regulated data into KiroCrew.
Treat anything you send as potentially logged or processed by the
configured model provider.
👻 Dashboard: ssh -NL 5476:localhost:5476 <EC2_HOST>
then open http://localhost:5476?token=<GATEWAY_TOKEN>
👻 Run 24/7: see docs/remote-desktop-setup.md for systemd service setup
👻 Headless remote session — skipping browser auto-open
ss -ltnp | grep ':5476'
The Gateway was listening on 127.0.0.1:5476. I will use the Dashboard URL in the form http://localhost:5476?token=<GATEWAY_TOKEN> shown at startup later.
ssh -N -L 45476:127.0.0.1:5476 ec2-user@<EC2_HOST>
The local 45476 is a free port I used during verification. If 45476 is in use, substitute any free port, and change the port number in the URL opened in the browser to the same value.
With the SSH tunnel established, change the port number of the Dashboard URL to 45476. Open http://localhost:45476?token=<GATEWAY_TOKEN> in the local browser.

I was able to display the Dashboard in the local browser.
Running a Chat from the Dashboard
I opened the model selection UI in a session.

I was able to choose from the list of available models.
I asked the agent in the session about the OS and architecture it was running on.

The response showed Amazon Linux 2023 and x86_64.
I asked it to check the caller's credentials with STS.

I was able to confirm the STS caller using the temporary credentials of the instance profile.
Summary
I started Kiro Crew on an Amazon Linux 2023 EC2 instance and was able to use the Dashboard from a local browser via an SSH tunnel.
By installing Kiro Crew on EC2, you can treat the EC2 instance as a sandbox dedicated to the agent. The impact of commands executed by the agent stays within the instance and does not reach your local machine. When you are done trying it out, you can clean up the whole environment just by deleting the stacks. Even when having the agent operate AWS APIs, you can use the temporary credentials of the instance profile as they are.
For light workloads such as chatting from the Dashboard, Kiro Crew ran on a t3a.medium. If you want to try Kiro Crew first, give running it on EC2 a try.
References
- https://kiro.dev/blog/introducing-kiro-crew/
- https://kiro.dev/docs/crew/
- https://github.com/kirodotdev/KiroCrew
- https://www.infoworld.com/article/4204961/awss-kiro-crew-aims-to-turn-ai-coding-agents-into-autonomous-engineering-teams.html
Reproduction Information
kirocrew-iam.yaml
AWSTemplateFormatVersion: '2010-09-09'
Description: KiroCrew IAM - EC2 Instance Profile with SSM access
Resources:
KiroCrewRole:
Type: AWS::IAM::Role
Properties:
RoleName: !Sub '${AWS::StackName}-role'
AssumeRolePolicyDocument:
Version: '2012-10-17'
Statement:
- Effect: Allow
Principal:
Service: ec2.amazonaws.com
Action: sts:AssumeRole
ManagedPolicyArns:
- arn:aws:iam::aws:policy/AmazonSSMManagedInstanceCore
KiroCrewInstanceProfile:
Type: AWS::IAM::InstanceProfile
Properties:
Roles:
- !Ref KiroCrewRole
Outputs:
InstanceProfileName:
Value: !Ref KiroCrewInstanceProfile
Export:
Name: !Sub '${AWS::StackName}-instance-profile'
RoleName:
Value: !Ref KiroCrewRole
Export:
Name: !Sub '${AWS::StackName}-role'
kirocrew-ec2.yaml
AWSTemplateFormatVersion: '2010-09-09'
Description: KiroCrew EC2 - x86_64 / AL2023 / kiro-cli + kirocrew auto-install
Parameters:
KeyName:
Type: AWS::EC2::KeyPair::KeyName
Default: m4mini
Description: EC2 キーペア名
IamStackName:
Type: String
Default: kirocrew-iam
Description: IAM スタック名(InstanceProfile のエクスポート元)
Resources:
KiroCrewSG:
Type: AWS::EC2::SecurityGroup
Properties:
GroupDescription: !Sub '${AWS::StackName} - SSH only'
SecurityGroupIngress:
- IpProtocol: tcp
FromPort: 22
ToPort: 22
CidrIp: 0.0.0.0/0
KiroCrewInstance:
Type: AWS::EC2::Instance
DeletionPolicy: Delete
UpdateReplacePolicy: Delete
Properties:
InstanceType: t3a.medium
ImageId: !Sub '{{resolve:ssm:/aws/service/ami-amazon-linux-latest/al2023-ami-kernel-default-x86_64}}'
KeyName: !Ref KeyName
IamInstanceProfile: !ImportValue
Fn::Sub: '${IamStackName}-instance-profile'
SecurityGroupIds:
- !Ref KiroCrewSG
MetadataOptions:
HttpTokens: required
HttpEndpoint: enabled
BlockDeviceMappings:
- DeviceName: /dev/xvda
Ebs:
VolumeSize: 20
VolumeType: gp3
Encrypted: true
DeleteOnTermination: true
Tags:
- Key: Name
Value: !Sub '${AWS::StackName}'
UserData:
Fn::Base64: !Sub |
#!/bin/bash
set -euo pipefail
LOG="/var/log/kirocrew-setup.log"
exec > >(tee -a "$LOG") 2>&1
echo "===== KiroCrew setup started: $(date) ====="
# ---- 1. /tmp 拡張 ----
mount -o remount,size=2G /tmp
echo "[1/6] /tmp 拡張: $(df -h /tmp | tail -1 | awk '{print $4}') 空き"
# ---- 2. スワップ 1GB ----
fallocate -l 1G /swapfile
chmod 600 /swapfile
mkswap /swapfile
swapon /swapfile
echo '/swapfile none swap sw 0 0' >> /etc/fstab
echo "[2/6] スワップ完了: $(swapon --show --noheadings)"
# ---- 3. システムパッケージ ----
dnf install -y python3.11 python3.11-pip nodejs npm git unzip
echo "[3/6] パッケージ完了"
# ---- 4. kiro-cli インストール (ec2-user) ----
sudo -u ec2-user bash << 'EOF'
set -euo pipefail
export HOME=/home/ec2-user
echo "[4/6] kiro-cli インストール中..."
KIRO_CLI_SKIP_SETUP=1 bash <(curl -fsSL https://cli.kiro.dev/install) <<< "y"
echo "[4/6] kiro-cli 完了: $(ls $HOME/.local/bin/kiro* | tr '\n' ' ')"
EOF
# ---- 5. kirocrew インストール (ec2-user) ----
sudo -u ec2-user bash << 'EOF'
set -euo pipefail
export HOME=/home/ec2-user
echo "[5/6] kirocrew インストール中..."
curl -fsSL https://download.crew.kiro.dev/cli.sh | sh
echo "[5/6] kirocrew 完了: $($HOME/.local/bin/kirocrew --version 2>/dev/null || echo installed)"
EOF
# ---- 6. PATH (.bashrc) ----
sudo -u ec2-user bash -c \
"grep -q '\.local/bin' ~/.bashrc || echo 'export PATH=\"\$HOME/.local/bin:\$PATH\"' >> ~/.bashrc"
echo "[6/6] .bashrc 更新完了"
echo "===== KiroCrew setup finished: $(date) ====="
echo "KIROCREW_SETUP_DONE" > /tmp/kirocrew-setup-done
Outputs:
InstanceId:
Value: !Ref KiroCrewInstance
PublicIp:
Value: !GetAtt KiroCrewInstance.PublicIp
SSMConnect:
Value: !Sub 'aws ssm start-session --target ${KiroCrewInstance} --region ${AWS::Region}'
SSHConnect:
Value: !Sub 'ssh -i ~/.ssh/${KeyName}.pem ec2-user@${KiroCrewInstance.PublicIp}'
NextStep:
Value: 'kirocrew setup && kirocrew config set agent.allow_unsandboxed_exec true && kirocrew gateway'
