Architecture canvas

Every workload, database and queue, on one canvas.

The canvas is not a diagram you maintain. It is read from the infrastructure Kapten provisions inside your account — so it is correct by construction, and editing a node changes the infrastructure rather than a picture of it.

Read from
Your AWS account, live
Objects
Services, data stores, queues, routes
Interaction
Pan, zoom, inspect, connect
Underneath
Kubernetes, Helm, Terraform
apiDeployment · 2 pods
webDeployment · 3 podsDeploying
workerDeployment · 1 pod
nightly-syncEventBridge → LambdaSleeping
postgresManaged Postgres · Multi-AZ
redisManaged Redis · cache tier
assetsS3 + CloudFront
jobsManaged queue · standard
productiondeploying into acme-production· ••••4021
100%

What it gives you

A map that cannot be out of date.

Architecture diagrams rot because they are documents. This one is a read of the account, so the only way for it to be wrong is for production to be wrong.

  • Read, never drawn

    The graph is derived from what is actually running in your VPC. There is no diagram file to keep updated, and no way for it to quietly drift from production.

  • Inspect any node

    Open a service and get its image, its replica count, its environment bindings and the deployment that put it there — the same inspector the dashboard uses.

  • Edges are real routes

    A line between two nodes means a security group rule and an internal DNS name that exist. Draw one and it is provisioned; remove it and it is revoked.

  • Data stores are first class

    Managed Postgres, Redis and queues are nodes with their own health and their own connection strings, not boxes drawn around something the diagram cannot see.

  • Health rides on shape

    Status is carried by the node silhouette as well as its colour, so a degraded service is legible at a glance — and stays legible to the ~8% of men who lose the green/amber/red axis.

  • One canvas per environment

    Production, staging and every preview environment get their own. Nobody reasons about staging while looking at a picture of production.

Ready when you are

Your cloud. Our ergonomics.

Connect a cloud account and watch a production-grade environment come up in your own VPC. If it isn’t serving traffic in fifteen minutes, we want to hear why.