Minimum Requirements

Kudashai ships in two forms. Kudashai Web is a single server binary (also available as a Helm chart) that embeds the web UI; you open it in a browser. Kudashai Desktop packages the same engine with a native window for macOS, Windows, and Linux, and adds desktop-only features such as port forwarding and automatic updates. Both connect to your clusters directly, so the machine running Kudashai must be able to reach every Kubernetes API server you want to manage.

Kudashai Web

Server host

	Minimum	Recommended
CPU	1 vCPU	2 vCPU or more
Memory	1 GB available for Kudashai	2 GB or more; add memory for large or many clusters
Disk	500 MB free for the binary, the local database, and kubeconfigs	SSD-backed storage
Operating system	Linux (amd64, arm64, i386), macOS (Intel, Apple Silicon), Windows (amd64, arm64, i386), FreeBSD	Linux amd64 for servers

Memory usage grows with the number of Kubernetes objects Kudashai watches. For every cluster you open, Kudashai keeps an in-memory copy of each resource kind you view (pods, deployments, events, and so on) so that the UI updates in real time. A cluster with tens of thousands of pods needs several hundred MB on its own. Size the host for the largest cluster you manage rather than for the number of clusters.

The Helm chart requests and limits 100m CPU and 256Mi memory by default. Raise resources.limits.memory in your values file before pointing Kudashai at production-sized clusters.

Kubernetes clusters

• Kubernetes 1.23 or newer. Kudashai reads the stable apps/v1, batch/v1, autoscaling/v2, policy/v1, networking.k8s.io/v1, and discovery.k8s.io/v1 APIs; autoscaling/v2 is the youngest of these and became stable in 1.23.
• A kubeconfig, bearer token, or client certificate with permission to list and watch the resources you want to see. Cluster-wide read access gives the fullest picture; namespace-scoped credentials work but hide cluster-scoped resources such as nodes and persistent volumes.
• Optional: metrics-server in the cluster for CPU and memory usage views. Without it Kudashai still works but shows no live usage figures.
• Reachability: the Kudashai host must reach each cluster's API server (TCP port 6443 on most distributions). Clusters behind a bastion can be reached through Kudashai's built-in SSH tunnel, which needs SSH access (TCP port 22) to the bastion host.

Database (Platform mode only)

Standalone modes store everything in a local SQLite file under ~/.kudashai and need no database server. Platform mode, which enables multi-user login and role-based access control, needs:

• PostgreSQL 13 or newer (the schema uses gen_random_uuid() and JSONB). The bundled deployment manifests use PostgreSQL 16.
• A database and a role that owns it. Kudashai creates and migrates its tables itself.
• Network access from the Kudashai host to PostgreSQL (TCP port 5432 by default).

Browser

Any current version of Chrome, Edge, Firefox, or Safari. The UI targets ES2020 and streams live updates over Server-Sent Events, so browsers released before 2020 are not supported.

For anything other than http://localhost, serve Kudashai over HTTPS (--certFile and --keyFile, or a TLS-terminating reverse proxy). Over plain HTTP/1.1 a browser opens at most six connections per host, and every open resource view holds one live stream, so views stop updating once that limit is reached. HTTPS enables HTTP/2, which removes the limit.

Network

Direction	Destination	Purpose
Inbound	TCP 7080 (default; change with --listen)	Users reaching the UI and API
Outbound	Kubernetes API servers	Cluster management
Outbound	PostgreSQL (Platform mode)	Shared database
Outbound	AI provider endpoints (optional)	AI features, only when a provider is configured. Self-hosted providers are supported, so no internet access is required
Outbound	SSH bastion hosts (optional)	Tunnelled clusters

Kudashai Desktop

Kudashai Desktop runs the same engine as Kudashai Web inside the application, with the UI rendered by the operating system's built-in web view (WebKit on macOS, WebView2 on Windows, WebKitGTK on Linux). Everything in the Kubernetes clusters and Network sections above applies to the desktop application as well.

Operating system

Platform	Requirement
macOS	macOS 10.13 (High Sierra) or newer. Intel and Apple Silicon builds are available
Windows	Windows 10 or 11, 64-bit (amd64 or arm64), with the Microsoft Edge WebView2 Runtime. Windows 11 and up-to-date Windows 10 already include it; otherwise the application prompts to download it on first launch
Linux	64-bit (amd64 or arm64) distribution with GTK 3 and libwebkit2gtk-4.0, for example Ubuntu 22.04 or Debian 12

Hardware

	Minimum	Recommended
CPU	64-bit dual-core	Quad-core
Memory	4 GB system memory, with about 1 GB free for Kudashai	8 GB or more when managing large clusters
Disk	500 MB free in the home directory for the local database, kubeconfigs, and staged updates	SSD

The same sizing rule as the server applies: memory is driven by the size of the clusters you open, plus the web view, which costs about as much as one browser tab.

License and network

• A Kudashai Desktop license key from the AwanIO Enterprise Portal. The key is entered in the application under Settings > License.
• Outbound HTTPS (TCP 443) to enterprise-portal.awan.io for license activation, the daily license check, and application updates. A cached license keeps the application working while offline, within its grace period.
• Local data lives under ~/.kudashai on macOS and Linux and under the user profile folder on Windows.
• Optional: PostgreSQL 13 or newer if you switch the desktop application to a shared database under Settings > Database, which turns it into a Platform-mode client.

Next Steps

• Installation — download and run Kudashai
• Getting Started — runtime modes and first login
• Multi-cluster — connecting several clusters