What importing does (and does not do), the four ways to view an estate, and what a resource's detail tabs offer.
Tracking resources
The Resources screen is where most of the day goes. It lists the infrastructure this platform is tracking for the selected project and environment.
Importing changes nothing in your cloud
Importing is a read. It asks the connection what exists, and records the ones you pick so the platform can watch, cost and manage them. It creates nothing, renames nothing and deletes nothing.
Untracking is the same in reverse: it removes the platform's record and leaves the real resource running. That is why the confirmation says so — untracking a database is not deleting a database.
A resource that was deleted in the cloud but is still tracked here shows as such, rather than disappearing: the platform does not quietly forget things on your behalf.
Four views of the same list
| View | Answers |
|---|---|
| Table | What is here, with everything sortable and filterable |
| Cards | The same, laid out for a narrow screen |
| Topology | Who deploys what — grouped by connection, then by type |
| Architecture | What can actually reach what — see Architecture diagrams |
Table and Topology both support multi-select, and the bulk toolbar appears once something is checked: reassign a connection, restart App Runner or ECS services, untrack. Below a phone breakpoint the table becomes cards automatically, and the cards carry the same checkbox — a forced layout change never takes a capability away.
Catalog metadata
Every tracked resource has a Catalog tab for the facts the provider does not hold: an owner, a repository URL, docs, tags. These are yours, filterable from the screen's own search, and they are what makes an estate somebody else built comprehensible six months later.
The Catalog tab is also where per-resource alert mutes live — see Alerts.
The Provision resource action is different from Import: it creates real infrastructure, runs as a background job, and tracks the result. It can start from a reusable Blueprint or from the resource form directly. The confirmation names that cloud resources and charges may be created before anything is queued.
Per-resource management
Clicking a resource opens its detail panel. What is offered depends on what it is, but the pattern is consistent: an Overview, a Configuration group, and then whatever that resource type can actually do. The complete matrix is in the Resource type reference. A few worth knowing about:
RDS and ElastiCache get Connection, SSH tunnel and Query tabs — where the endpoint is, how to reach it, and a console to run against it. Redis command classification is enforced by the platform, so a read-only session refuses a write rather than sending it.
OpenSearch domains get Indices, Query, Access and Dashboards tabs. The Access tab is the one to open first for a VPC domain: it reports how the platform can reach the domain and can find the load balancer in front of it.
Lambda functions get environment variables, image, network and cross-account compare/sync, plus logs, schedules, usage history and client-side encryption of individual variables.
ECS services get environment variables with secret support, and compare/sync against another service.
S3, SES, API Gateway, CloudFront, EventBridge and security groups get purpose-built inventory, configuration or operational tabs instead of a generic JSON view.
Azure App Service, Container Apps, SQL Database and Storage Accounts, GCP Cloud Run, Cloud SQL, Compute, VPC, Storage, Firestore, Pub/Sub, Redis and Functions, Atlas, and Cloudflare resources open provider-specific panels when that connector supports them.
Parameter Store hierarchies get a scan for values with trailing whitespace and for values that look like secrets stored in plaintext.
Status, and what "unhealthy" means
Status is synced on a schedule and classified into tones — healthy, degraded, failing. Two things are deliberately separate:
Sustained unhealthiness raises a notification after several consecutive bad samples.
Flapping counts state transitions in a window, so a resource that is steadily down does not qualify. They are complementary, not two names for one thing.
An ECS service is the case where status alone lies: ECS reports ACTIVE for a service whose every
task is crash-looping. That is what the service-unhealthy alert exists for — it compares running
capacity against desired.