CloudConsole

Ask CloudConsole: the AI chat about your AWS accounts

Ask questions about your AWS accounts and GitHub code in a chat; Claude runs on Amazon Bedrock in your own account, reads with your permissions and only proposes changes. Pro.

Tap the sparkle button (on a phone, the small button at the edge of the tab bar; on an iPad or a wide window it opens a panel beside the screen you're on) and ask a question about your accounts: “Why is my bill high?”, “What changed in production in the last hour?”, “Which repo defines orders-worker?”. You can also start from Why is my bill high? on Billing, from More → Ask about this on any resource (the resource, account and region go along as context) or from Continue in chat after Explain.

Ask CloudConsole answering “Why is my bill high?” across two accounts: NAT gateway data processing and an idle load balancer, with numbered sources
An answer with its sources
A proposed pull request: the diff raising orders-worker to 4 tasks, with Review & open PR and Dismiss
A proposed change waits for you

How it answers

Claude decides what to look up, the app runs those lookups on your phone, and Claude answers from what came back. Each lookup is shown as a line in the chat (tap it to see exactly what was asked), and facts in the answer carry numbered sources such as S1 that open the screen they came from. A question can take several rounds of lookups, up to eight; Stop ends it at the next step. Follow-up questions keep the context.

Which accounts and regions

The pill at the top of the chat chooses the scope, as on Home: this account or several, one region or many. The chat says which account each number comes from, and an account that is signed out or refuses a call is reported without stopping the answer. Claude itself always runs in the account you are using, in the Bedrock region set in Explain & Ask (AI); the header says so, for example “Claude runs in acme-prod / eu-west-2; it reads data from the 3 accounts you chose”.

What it can read

  • Costs: Cost Explorer by service, usage type, region, linked account or tag, compared with the previous period, the daily trend and forecast, Cost Anomaly Detection, the cost of each resource and the waste finder's unused resources. Cost Explorer charges US$0.01 a request, as everywhere in the app.
  • Resources: any resource type through AWS Cloud Control, a single resource's settings (passwords, secrets and environment variable values removed), resources by tag, and the app's own search index and Home findings.
  • Metrics, alarms and changes: CloudWatch metrics, alarms and their history, and CloudTrail write events.
  • Logs: recent errors in a log group, and Logs Insights queries. A query over up to one hour and one log group runs straight away; a wider one appears as a card with the estimated data scanned and cost, and runs only when you tap it.
  • Security: GuardDuty, Security Hub, Inspector and AWS Config findings, the IAM credential report and “who can do this?” (the IAM policy simulator).
  • GitHub (if connected): your watched repositories' code search, files, recent commits, pull requests and workflow runs, and which stack or file defines a resource. Turn AI can read off for a repository in the GitHub settings to keep it out of the chat.

Changes are only proposed

The chat never changes anything by itself. It can propose scaling an ECS service, stopping or starting an EC2 instance, turning an alarm's actions on or off, setting a log group's retention, opening a pull request (one change in one file, shown as a diff) or opening a GitHub issue. Each appears as a card; nothing happens unless you tap it, and then it goes through your guardrails and the activity log like any other change. Pull requests and issues are only offered on repositories your GitHub token can write to; for the others the chat says the repository is read-only.

Set it up

The chat uses the same setup as AI Explain: model access to a Claude model in Amazon Bedrock, bedrock:InvokeModel for the role you sign in with, and Account → Integrations → Explain & Ask (AI) turned on. For the lookups it uses the read permissions your roles already have; ReadOnlyAccess covers them. With a narrower role, these are the actions behind each kind of lookup:

{
  "Effect": "Allow",
  "Action": [
    "bedrock:InvokeModel",
    "ce:GetCostAndUsage", "ce:GetCostForecast", "ce:GetAnomalies",
    "cloudformation:ListResources", "cloudformation:GetResource",
    "tag:GetResources",
    "cloudwatch:GetMetricData", "cloudwatch:DescribeAlarms", "cloudwatch:DescribeAlarmHistory",
    "cloudtrail:LookupEvents",
    "logs:FilterLogEvents", "logs:StartQuery", "logs:GetQueryResults",
    "guardduty:ListDetectors", "guardduty:ListFindings", "guardduty:GetFindings",
    "securityhub:GetFindings", "inspector2:ListFindings",
    "config:DescribeConfigRules", "config:DescribeComplianceByConfigRule",
    "iam:GetCredentialReport", "iam:SimulatePrincipalPolicy"
  ],
  "Resource": "*"
}

Cloud Control reads also need the read permissions of the resource type you ask about (for example ec2:DescribeInstances). A lookup your role can't make is reported in the chat and the rest still runs.

Privacy and cost

Your questions and what the lookups return go from your phone to Amazon Bedrock in your own AWS account, signed with that account's credentials; that can include log lines and code you asked about. Nothing passes through CloudConcur or any other AI service, and Bedrock doesn't store prompts or train on them by default. Chats are kept on your phone (the last 30, under Past chats) and nowhere else. The model use is billed by AWS to your account at Bedrock's prices; the footer of each chat shows the tokens used and an estimate in US dollars. The chat is part of Pro; in the demo it works without Pro, on sample data.

Questions

Can the chat delete or change my resources?

Not by itself. It can only propose a small set of changes as cards, and each one runs only when you tap it, through your usual confirmation and the activity log.

Does CloudConcur see my chats?

No. The chat runs between your phone, AWS (Bedrock and the services it reads) and GitHub if you connected it. Chats are stored on your phone only.

Why did it ask me before running a Logs Insights query?

Logs Insights is billed by the data it scans. Queries over more than an hour or one log group show an estimate first and run only when you tap them.

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.