A new open-source project published on GitHub, k10s, offers a Kubernetes terminal user interface that can be operated with a mouse as well as the keyboard. The tool is written in Go and ships as a single binary. Its author presents it as a more approachable companion to k9s, the widely used terminal dashboard for managing clusters, and adds instant search, an AI assistant that understands the selected object, and a demo mode that runs without any cluster at all.
Why a clickable terminal dashboard matters
Many engineers keep a cluster dashboard open for most of the working day, often during an incident when speed matters most. According to the project's README on GitHub, the two qualities that count are how quickly the dashboard opens and how little a user needs to memorise before getting useful work done.
Terminal tools for Kubernetes have long rewarded people who learn a set of single-letter shortcuts. That works well for experienced operators but can slow down colleagues who only check a cluster occasionally. k10s tries to close that gap by making the interface discoverable: the actions that apply to the selected resource appear in their own pane, and each one can be clicked or triggered by pressing the letter shown beside it.
What k10s offers
The README highlights several features that set the tool apart from a plain command-line workflow:
- Mouse support: rows, panes, namespace pickers and scrolling all respond to clicks.
- One search box: ctrl+p searches resource kinds and individual objects together, with no prefix syntax to learn.
- Logs, exec and port-forward: these actions are available from the same interface where the user is already looking.
- Themes: several built-in colour schemes are available, with tokyo-night set as the default, and custom palettes can be added as YAML files.
- Plugins: users can add scoped shortcuts in a plugins file that follows the same format as k9s, and those shortcuts receive the selected object, namespace and context.
- Self-update: a /update command installs the newest release over the running binary, with checksum verification.
The project also states that startup is deliberately light. Rather than watching every resource type at launch, k10s starts informers lazily, one resource kind at a time, so opening the pods view only watches pods. The README links to a set of regression guards that explain how this behaviour is tested.
Resource coverage is broad. The sidebar groups 30 kinds across workloads, networking, configuration, storage, RBAC, cluster resources and custom resources. Custom resource definitions are discovered automatically, without manual configuration. Groups can be folded to reduce noise, and a folded group does not trigger requests to the cluster for counting.
The AI assistant and data privacy
The most distinctive addition is an AI prompt, opened with ctrl+a. The current context, namespace, resource kind and selected object are included in the prompt, so a question such as why is this pod unhealthy refers to the pod on screen. Users bring their own API key and can connect to any OpenAI-compatible endpoint, including local servers such as Ollama and LM Studio, or to Anthropic.
The README is explicit about data flow. No cluster data leaves the user's machine unless the user presses enter in AI mode. The only other network call the tool makes on its own is a daily update check that requests a version number from GitHub.
There is one caveat worth noting. The project's documentation says the AI API key is stored as plain text in the configuration file, with restricted file permissions, and is masked within the interface. Teams with strict secrets policies may prefer to leave that field empty, in which case AI mode stays off.
Trying it without a cluster
Demo mode and installation
k10s includes an offline demo backend with sample data, including a pod in a CrashLoopBackOff state, so new users can explore the interface without connecting to a real environment. Running k10s demo opens the demo context, and the header displays a DEMO marker to prevent confusion with a live cluster.
Installation is supported on macOS and Linux, on both amd64 and arm64 processors. The installer checks the binary's sha256 hash against the release manifest. Prebuilt binaries are also published for Windows, although the installer directs Windows users to the release page instead. Plain k10s reads the same kubeconfig file that kubectl uses, so it shows only what the current context can reach.
How it compares with k9s
The project is open about its relationship to k9s, which it credits as mature, far broader in scope and extensible through plugins. The README says k9s suits users who want the largest feature set and an established plugin ecosystem. k10s, by contrast, is positioned as narrower but easier to hand to a colleague who has never used a terminal dashboard before.
That positioning suggests k10s is best understood as a complement rather than a replacement, at least for now. Teams already invested in k9s workflows will have less reason to switch, while those onboarding newcomers to Kubernetes may find the mouse support and visible action panes more intuitive.
Caveats and what to watch
The author states plainly that the code has not yet been smoke-tested against a production cluster. Live features are implemented against the official client-go and kubectl APIs and covered by tests that use fake clientsets, but the README recommends trying the tool on a local test cluster such as kind first. Known limitations are listed in the project's roadmap, along with planned work such as click-to-sort columns, multi-select bulk deletion, searching within log streams and saved views.
The project is licensed under Apache-2.0, the same licence used by Kubernetes, Helm and k9s, which makes it straightforward for companies to evaluate, fork or adapt internally.
For platform teams, k10s is worth a short trial on a non-production cluster. Its real test will be whether it earns a place next to established tools once it has been used under production pressure.
Related: Open-Source n8n Bot Calculates Czech Tax on Broker Investments · Software License Renewal Tracker Sends Timely Reminders
Comments (0)