Introduction

InfraCanvas is a real-time map of your infrastructure. Connect a machine and within seconds you see its hosts, Docker containers, and Kubernetes resources as a single live topology — health, relationships, metrics, and logs included. Then you act on it: open a terminal, scale a deployment, edit a manifest, or ask Claude — all from the browser. No YAML to write, no Helm, no SSH tunnels, no inbound ports.

How it works

InfraCanvas has two pieces:

• The agent — a single lightweight Go binary you run on each machine. It discovers the host, Docker, Kubernetes, and LXD/Incus locally and streams a graph to the cloud over one outbound WebSocket. It never opens an inbound port.
• The dashboard — cloud.infracanvas.app, where the live canvas, terminals, actions, metrics, and alerts live.

your VM ──(outbound WSS)──▶ relay ──▶ dashboard (your browser)
  └─ agent: host · docker · kubernetes · lxd

The agent reads only infrastructure metadata using the same interfaces you already have — the Docker socket and your kubeconfig. Secret values never leave the machine. See Security & data handling for the full data flow.

What you can do

See everything, live

• Zero-config discovery. Point the agent at a box and get a full topology — hosts, containers, images, networks, volumes, and the entire Kubernetes object graph (nodes, namespaces, workloads, pods, services, ingresses).
• A live canvas, not a snapshot. The graph updates as containers start, pods reschedule, and nodes come and go — with health rolled up at every level.
• Multiple machines, one view. Map a whole fleet and navigate between machines, clusters, and namespaces from the same canvas.

Operate from the canvas

• Web terminals. Open an interactive shell on a host or inside a container — no SSH keys, no inbound ports. Sessions stream over the same WebSocket.
• One-click actions. Restart a container, scale a deployment, restart a rollout, restart a systemd service, delete a pod, and more — scoped to each resource type.
• Edit manifests inline. View any Kubernetes object's live YAML, edit it, and apply — with a reset to the live version one click away.
• Stream logs. Tail container, pod, host, and service logs live, then export them as .log or .txt.

See Actions & terminals for the full list.

Understand trends

• Metrics & history. CPU (overall and per-core), memory, disk usage and throughput, load average, and network — collected continuously and charted as trends per machine and across your fleet. See Metrics & history.

Automate & integrate

• Ask Claude. Connect the MCP server and query or operate your infrastructure in natural language from Claude or any MCP client.
• AI assistant. Built-in assistant for investigating health and running guided operations from the dashboard.
• Alerts. Route health and threshold alerts to Email, Slack, and PagerDuty.
• Batch operations. Run the same action across many resources at once.

Govern access

• RBAC & approval gates. Control who can view and who can act, with optional approval gates on sensitive operations.
• Audit log. Every action is recorded with the acting user's email, the target resource, and a timestamp.
• Multi-account. One machine can stream to several organizations at once without disconnecting any of them. See Connect a VM to multiple accounts.

Get started