Permissions for changes (Pro actions)
Which IAM actions each button in the app needs, with a policy to copy.
Changes need two things: CloudConsole Pro, and an AWS role that allows the action. Where each button is: What you can change from your phone. Pro doesn't grant any AWS permission. If AWS denies a change, nothing is changed and the app shows the exact action and a minimal policy to ask for.
Every change the app can make
| Service | Buttons in the app | IAM actions |
|---|---|---|
| EC2 | Start, Stop, Reboot, Terminate; under More: type, snapshot, AMI…: Change instance type, Create EBS snapshot, Create image (AMI), Termination protection | ec2:StartInstancesec2:StopInstancesec2:RebootInstancesec2:TerminateInstancesec2:ModifyInstanceAttributeec2:CreateSnapshotec2:CreateImage |
| Waste finder | Snapshot & delete, Delete (volume, NAT gateway, load balancer), Delete snapshot, Release (Elastic IP), Set 30-day retention (log group) | ec2:DeleteVolumeec2:CreateSnapshotec2:DeleteSnapshotec2:ReleaseAddressec2:DeleteNatGatewayelasticloadbalancing:DeleteLoadBalancerlogs:PutRetentionPolicy |
| Monthly cost card | Change to gp3 (a gp2 EBS volume); a Graviton class on an RDS database opens RDS Change class | ec2:ModifyVolume |
| ECS | Force new deployment, Scale, Roll back (to the previous revision), Turn on / Turn off circuit breaker, Stop task | ecs:UpdateServiceecs:StopTask |
| EC2 Auto Scaling | Set desired capacity, Min & max, Suspend processes (and resume them), Start instance refresh, Cancel instance refresh | autoscaling:SetDesiredCapacityautoscaling:UpdateAutoScalingGroupautoscaling:SuspendProcessesautoscaling:ResumeProcessesautoscaling:StartInstanceRefreshautoscaling:CancelInstanceRefresh |
| Lambda | Invoke, Memory & timeout, Throttle (emergency stop), Unthrottle, Edit environment variables, Publish version, Change version (of an alias), Reserved concurrency | lambda:InvokeFunctionlambda:UpdateFunctionConfigurationlambda:PublishVersionlambda:UpdateAliaslambda:PutFunctionConcurrencylambda:DeleteFunctionConcurrency |
| S3 | Share link (presigned URL), Delete (an object), Restore (an archived object), Turn on versioning / Resume versioning, Block all public access | s3:GetObjects3:DeleteObjects3:RestoreObjects3:PutBucketVersionings3:PutBucketPublicAccessBlock |
| DynamoDB | Edit item, Delete item, Switch to on-demand / Switch to provisioned, Turn on point-in-time recovery | dynamodb:PutItemdynamodb:DeleteItemdynamodb:UpdateTabledynamodb:UpdateContinuousBackups |
| RDS | Start, Stop, Reboot (or Reboot with failover on Multi-AZ), Snapshot, Change class, Turn on / Turn off deletion protection | rds:StartDBInstancerds:StopDBInstancerds:RebootDBInstancerds:CreateDBSnapshotrds:ModifyDBInstance |
| SQS | Start redrive, Stop (a redrive), Purge, Peek messages, Send test message | sqs:StartMessageMoveTasksqs:CancelMessageMoveTasksqs:PurgeQueuesqs:ReceiveMessagesqs:SendMessage |
| Step Functions | Start execution, Run again, Stop, Redrive | states:StartExecutionstates:StopExecutionstates:RedriveExecution |
| CodePipeline | Approve / Reject, Release change, Retry failed actions (Retry on Home), Stop execution | codepipeline:PutApprovalResultcodepipeline:StartPipelineExecutioncodepipeline:RetryStageExecutioncodepipeline:StopPipelineExecution |
| CloudFormation | Cancel update, Continue update rollback | cloudformation:CancelUpdateStackcloudformation:ContinueUpdateRollback |
| CloudWatch alarms | Mute notifications / Unmute (Mute on Home), Set state to OK, Create alarm (from a metric chart) | cloudwatch:DisableAlarmActionscloudwatch:EnableAlarmActionscloudwatch:SetAlarmStatecloudwatch:PutMetricAlarm |
| CloudWatch Logs | Retention (on a log group) | logs:PutRetentionPolicylogs:DeleteRetentionPolicy |
| EventBridge | Enable rule / Disable rule | events:EnableRuleevents:DisableRule |
| Secrets Manager | Rotate now, Update value | secretsmanager:RotateSecretsecretsmanager:PutSecretValue |
| Certificate Manager (ACM) | Request (a certificate), Resend validation email, Delete (a certificate nothing uses) | acm:RequestCertificateacm:ResendValidationEmailacm:DeleteCertificate |
| Systems Manager | Parameter Store: Create, Edit value, Delete; Run Command: Run (also from EC2 → More → Run command) | ssm:PutParameterssm:DeleteParameterssm:SendCommand |
| CodeBuild | Start build, Retry build (Retry on Home), Stop build | codebuild:StartBuildcodebuild:RetryBuildcodebuild:StopBuild |
| Route 53 | Edit a record's value and TTL (the pencil on a record), then follow the change until it is in sync | route53:ChangeResourceRecordSetsroute53:GetChange |
| IAM | Deactivate / Reactivate an access key | iam:UpdateAccessKey |
| CloudFront | Create invalidation | cloudfront:CreateInvalidation |
| Tags (any resource screen with a Tags section) | Add tag, Edit, Remove | tag:TagResourcestag:UntagResources |
| Cloud Control services (about 110) | Edit a property, Delete | cloudformation:UpdateResourcecloudformation:DeleteResource |
| Billing (AWS Budgets) | Create a budget alert | budgets:ModifyBudget |
| Tools: Service quotas | Request increase | servicequotas:RequestServiceQuotaIncrease |
A policy with all of them
Add it to a separate "operator" role, or alongside a read-only policy. "Resource": "*" is the widest form; narrow it to specific ARNs, or add conditions on tags, wherever you can.
{
"Version": "2012-10-17",
"Statement": [
{
"Sid": "CloudConsoleActions",
"Effect": "Allow",
"Action": [
"acm:DeleteCertificate",
"acm:RequestCertificate",
"acm:ResendValidationEmail",
"autoscaling:CancelInstanceRefresh",
"autoscaling:ResumeProcesses",
"autoscaling:SetDesiredCapacity",
"autoscaling:StartInstanceRefresh",
"autoscaling:SuspendProcesses",
"autoscaling:UpdateAutoScalingGroup",
"budgets:ModifyBudget",
"cloudformation:CancelUpdateStack",
"cloudformation:ContinueUpdateRollback",
"cloudformation:DeleteResource",
"cloudformation:UpdateResource",
"cloudfront:CreateInvalidation",
"cloudwatch:DisableAlarmActions",
"cloudwatch:EnableAlarmActions",
"cloudwatch:PutMetricAlarm",
"cloudwatch:SetAlarmState",
"codebuild:RetryBuild",
"codebuild:StartBuild",
"codebuild:StopBuild",
"codepipeline:PutApprovalResult",
"codepipeline:RetryStageExecution",
"codepipeline:StartPipelineExecution",
"codepipeline:StopPipelineExecution",
"dynamodb:DeleteItem",
"dynamodb:PutItem",
"dynamodb:UpdateContinuousBackups",
"dynamodb:UpdateTable",
"ec2:CreateImage",
"ec2:CreateSnapshot",
"ec2:DeleteNatGateway",
"ec2:DeleteSnapshot",
"ec2:DeleteVolume",
"ec2:ModifyInstanceAttribute",
"ec2:ModifyVolume",
"ec2:RebootInstances",
"ec2:ReleaseAddress",
"ec2:StartInstances",
"ec2:StopInstances",
"ec2:TerminateInstances",
"ecs:StopTask",
"ecs:UpdateService",
"elasticloadbalancing:DeleteLoadBalancer",
"events:DisableRule",
"events:EnableRule",
"iam:UpdateAccessKey",
"lambda:DeleteFunctionConcurrency",
"lambda:InvokeFunction",
"lambda:PublishVersion",
"lambda:PutFunctionConcurrency",
"lambda:UpdateAlias",
"lambda:UpdateFunctionConfiguration",
"logs:DeleteRetentionPolicy",
"logs:PutRetentionPolicy",
"rds:CreateDBSnapshot",
"rds:ModifyDBInstance",
"rds:RebootDBInstance",
"rds:StartDBInstance",
"rds:StopDBInstance",
"route53:ChangeResourceRecordSets",
"route53:GetChange",
"s3:DeleteObject",
"s3:GetObject",
"s3:PutBucketPublicAccessBlock",
"s3:PutBucketVersioning",
"s3:RestoreObject",
"secretsmanager:PutSecretValue",
"secretsmanager:RotateSecret",
"servicequotas:RequestServiceQuotaIncrease",
"sqs:CancelMessageMoveTask",
"sqs:PurgeQueue",
"sqs:ReceiveMessage",
"sqs:SendMessage",
"sqs:StartMessageMoveTask",
"ssm:DeleteParameter",
"ssm:PutParameter",
"ssm:SendCommand",
"states:RedriveExecution",
"states:StartExecution",
"states:StopExecution",
"tag:TagResources",
"tag:UntagResources"
],
"Resource": "*"
}
]
}
Notes
- SQS redrive moves messages from a dead-letter queue back to its source. AWS also requires permission to read and delete messages on the dead-letter queue and to send them to the destination queue (
sqs:ReceiveMessage,sqs:DeleteMessage,sqs:GetQueueAttributes,sqs:SendMessage). - S3 Share link makes a presigned URL on your phone; nothing is sent to AWS until someone opens it, and AWS then checks
s3:GetObjectfor you (the signer). It works for up to 7 days, or until your sign-in expires if you use SSO or a role, whichever is sooner. - DynamoDB Edit item / Delete item only go through if the attributes you saw are unchanged. Switching from provisioned to on-demand is allowed up to four times in any 24 hours; on-demand to provisioned at any time.
- RDS Change class lists the classes your engine version supports with
rds:DescribeOrderableDBInstanceOptions(read-only; without it you type the class). Applied immediately it means several minutes of downtime. - Peek at messages counts as a change because receiving a message hides it from consumers for the visibility timeout.
- Changing the type of a running EC2 instance runs three steps: stop, change the type, start. It needs
ec2:StopInstances,ec2:ModifyInstanceAttributeandec2:StartInstances. A stopped instance only needsec2:ModifyInstanceAttribute. - Create image (AMI) doesn't reboot the instance unless you switch that on. Copy SSM session command only copies
aws ssm start-sessionfor your computer; it changes nothing and is free. - ECS roll back deploys the newest active task definition revision older than the current one, after showing which container images change. Finding it uses
ecs:ListTaskDefinitionsandecs:DescribeTaskDefinition. - Lambda environment variables change
$LATESTonly. Published versions keep their values, so an alias that points at a version needs a new version (lambda:PublishVersion) and the alias moved (lambda:UpdateAlias). - Showing a secret's value calls
secretsmanager:GetSecretValue. It isn't a change, but it is recorded in CloudTrail. - Revealing a SecureString parameter calls
ssm:GetParameterwith decryption, and for a customer managed KMS key also needskms:Decrypton that key (AWS's defaultalias/aws/ssmkey needs nothing extra). Saving a SecureString with a customer managed key needskms:Encrypton it. - Run Command (
ssm:SendCommand) runs as root on the instance. Narrow it with the document (arn:aws:ssm:*:*:document/AWS-RunShellScript) and the instances' ARNs or tags. Reading the result usesssm:ListCommandInvocationsandssm:GetCommandInvocation. - CodeBuild Retry build reruns the same source version. For a build a pipeline started, retry the pipeline stage instead to carry the release on.
- Tags go through the Resource Groups Tagging API, which also needs the resource's own tag permission (for example
ecr:TagResource/ecr:UntagResource, orcloudwatch:TagResource). Tags whose key starts withaws:belong to AWS and can't be changed. - Cloud Control (Edit a property, Delete on the ~110 catalogue services) calls the service as you, so it also needs that service's own update or delete permission (for example
ecr:PutImageTagMutabilityorecr:DeleteRepository). The buttons only appear when the resource type's schema (cloudformation:DescribeType) has an update or delete handler, and only simple properties that can change in place are offered. - Create a budget alert: AWS Budgets has no separate CreateBudget permission; creating a budget and its notifications needs
budgets:ModifyBudget. - Your organisation's service control policies apply on top. If one blocks an action, no role change helps; the app says so.
Still stuck?
Email [email protected] with what you tried and any message the app showed. Don’t send access keys or secrets. We reply within two working days.