Every cloud you sign in to, in one place
Open Infra › Cloud providers and pick a cloud. AWS lists the profiles in your aws config, and you choose the ones AI Bridge may use. Azure lists the subscriptions your az sign-in can see. Google Cloud lists the projects your gcloud account can see, and opens any other by its ID. Inside each: an overview, a table for each kind of resource, a drawer for each one, and links to the cloud’s own console.
What runs in them
- AWS, across every region the account has turned on: EC2, EKS and Lambda; ECS and ECR; RDS and DynamoDB; EBS volumes and S3 buckets; VPCs, subnets, NAT gateways, Elastic IPs and security groups; load balancers, API Gateway, CloudFront and Route 53 records; SQS queues and SNS topics; IAM users, roles, groups and policies; and the CloudWatch alarms that are firing.
- Azure, a whole subscription or one resource group at a time: every resource it holds, grouped into compute, web and containers, data, network, security and governance. AKS clusters show the Kubernetes versions they can upgrade to, container registries their use against their limits, PostgreSQL flexible servers their databases and firewall rules, and VMs, AKS clusters and PostgreSQL servers their main metrics over the last 24 hours.
- Google Cloud, project by project: Compute Engine VMs, GKE clusters, Cloud Run services and Cloud SQL instances in every location, Cloud Storage buckets and whether the public can read them, and every other kind of resource Cloud Asset Inventory counts.
What needs a look comes first
Each overview opens on what needs you, in the cloud’s own terms:
- AWS: alarms that are firing, failed status checks, deprecated runtimes and databases without backups; then what could cost less, such as unattached volumes and idle Elastic IPs, and what is open to the internet, such as a security group open to any address.
- Azure: changes that failed, VMs stopped but still billed and registries with the admin user on, beside Azure Advisor’s recommendations.
- Google Cloud: VMs stopped with their disks still billed, clusters in error and Cloud SQL instances open to the internet or without backups, beside Google Cloud Recommender’s advice.
Costs, and the reads that cost
AWS shows Cost Explorer’s figures: this month by service, and each day since the first of last month. Azure shows Cost Management’s: this month so far, last month and each day since, and what each resource and resource group costs, with an AKS cluster’s node resource group counted under the cluster. Google Cloud has no API for costs, so its page says so and links to its billing reports.
Some reads cost money themselves. AWS charges for Cost Explorer, CloudWatch metrics and S3, SQS and SNS requests, and Google Cloud bills Cloud Monitoring and Cloud Storage reads past the free allowance. AI Bridge makes those only when you click, after a card that says how many requests it will make, the most they can cost and on whose bill. Nothing billed is ever read on its own.
Analyse with AI
Analyse with AI reads one account, subscription or project, or every one you tick across a cloud, and hands an AI a summary: what runs where, how busy it is, what needs a look and, where the cloud says, what it costs. The AI writes back a report of findings, and a finding about a resource opens it. It runs through the AI coding CLI you pick, Claude Code, Codex, Grok or OpenCode, on your own plan with that provider, and you pick the model.
Before anything is sent, you read exactly what goes and to which AI, with an estimate of its tokens. Untick Send resource names and the AI sees “Virtual machine 3” in place of a name, while your report still shows the real ones. Reports are kept on your Mac, the past ones a click away.
Your sign-in, as it is
There is nothing to paste. When AI Bridge reads, it asks your own aws, az or gcloud for credentials, keeps them in memory and calls the cloud from your Mac: nothing passes through AI Bridge’s servers. It never signs you in or out, switches your subscription or project, or changes your CLI’s settings. When a sign-in has expired, it shows you how to sign in again.
Clusters open in Kubernetes
An AKS cluster’s drawer has Open in Clusters, which takes you to it in AI Bridge’s Kubernetes view, with its pods, logs, YAML and Helm releases. If it is not in your kubeconfig yet, Add to Clusters opens a terminal with the az command typed in, and nothing runs until you press Return. EKS and GKE clusters show their versions and nodes on their cloud’s page, and open in Kubernetes like any cluster in your kubeconfig. No cluster there yet? Clusters offers to set one up with Assist, for AKS, EKS or GKE.
Beside the rest of the job
AI Bridge runs Claude Code, Codex, Grok and OpenCode side by side, each in its place. Around them it is your Kubernetes client, like Lens, and your database client for Postgres and MongoDB, like Postico, with Slack agents, an iPhone app and a browser extension that lets your agents work in your own browser. For anything the cloud pages don’t show, ask for a custom view: Assist or a CLI you pick builds it, and a command it reads from, such as an aws or gcloud command, runs only once you’ve approved it. See how it all fits together.
