SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI designk8m

k8m

k8m is a lightweight, cross-platform Mini Kubernetes AI Dashboard console tool designed specifically to simplify Kubernetes cluster management and operations. It builds its interface based on AMIS, uses kom as the Kubernetes API client, has built-in Qwen2.5-Coder-7B, and supports interaction with the deepseek-ai/DeepSeek-R1-Distill-Qwen-7B model. It can also connect to users' own private large models (including Ollama). k8m integrates multi-cluster management, intelligent analysis, real-time anomaly detection, MCP integration and permission connectivity, cluster inspection, Event forwarding, CRD management, Helm marketplace, Pod file and runtime management, API exposure, and other capabilities, and supports cluster types such as standard k8s, AWS EKS, k3s, kind, and k0s. The backend uses Golang, the frontend is based on Baidu AMIS, and all functions are integrated into a single executable file. It supports multiple architectures such as x86 and ARM, is compatible with Linux, macOS, and Windows, supports multiple databases such as SQLite, MySQL, and PostgreSQL, and can be deployed as a single file. The project is fully open source, allows free customization and extension, and can be used commercially. The default login username is k8m, and the password is k8m. After going live, the username and password must be changed and two-step verification enabled.

Open website

About this tool

k8m is an AI-driven Mini Kubernetes AI Dashboard lightweight console tool designed specifically to simplify cluster management. It is built based on AMIS and uses kom as the Kubernetes API client. Its design philosophy is "AI-driven, lightweight and efficient, simplifying the complex," helping developers and operations personnel get started quickly and easily manage Kubernetes clusters.

In terms of purpose: k8m is aimed at daily management and operations optimization of Kubernetes clusters, providing a graphical console, supporting multi-cluster management and permission control, and using AI capabilities to assist with resource interpretation, log analysis, and operations decision-making.

Main capabilities:

  • Miniaturized design: all functions are integrated into a single executable file, making deployment convenient and use simple.
  • Easy to use: a friendly user interface and intuitive operation flow, supporting cluster types such as standard k8s, AWS EKS, k3s, kind, and k0s.
  • High efficiency and performance: the backend is built with Golang, and the frontend is based on Baidu AMIS, ensuring high resource utilization and fast response speed.
  • AI-driven integration: based on ChatGPT, it implements selected-text explanation, resource guides, automatic translation of YAML attributes, interpretation of Describe information, AI log diagnosis, and recommended runtime commands, and integrates k8s-gpt functionality to achieve Chinese presentation.
  • Plugin-based functionality: feature functions are plugin-based and enabled on demand; if not enabled, they do not consume resources.
  • MCP integration: visual management of MCP, enabling large models to call Tools, with 49 built-in k8s multi-cluster MCP tools that can be combined to achieve more than a hundred cluster operations, and can be used as an MCP Server by other large-model software. It can record every MCP call in detail and supports mainstream mcp.so services.
  • MCP permission connectivity: multi-cluster management permissions are connected with MCP large-model invocation permissions. Whoever uses the large model executes MCP with their own permissions, avoiding unauthorized operations.
  • Multi-cluster management: automatically identifies InCluster mode used inside the cluster. After configuring the kubeconfig path, it automatically scans configuration files in the same-level directory and simultaneously registers and manages multiple clusters, supporting heartbeat detection and automatic reconnection.
  • Multi-cluster permission management: supports authorization for users and user groups, and can authorize by cluster, including three permissions: cluster read-only, Exec command, and cluster administrator. After authorizing a user group, users in the group all obtain the corresponding authorization. Supports setting namespace blacklists and whitelists.
  • Supports the latest k8s features: supports features such as APIGateway and OpenKruise.
  • Pod file management: in the file tree on the left side of the Console interface, the right-click menu supports browsing, editing, uploading, downloading, and deleting files inside a Pod.
  • Pod runtime management: supports viewing Pod logs in real time, downloading logs, and executing Shell commands directly inside a Pod. Supports Ctrl+F search, similar to grep -A -B highlighted search.
  • Open API: supports creating API KEYs, accessing from third-party external systems, and provides a swagger interface management page.
  • Cluster inspection support: supports scheduled inspection of multiple clusters and custom inspection rules, and supports lua script rules. Supports sending to DingTalk groups, WeChat groups, Feishu groups, and custom webhooks. Supports AI summaries.
  • k8s Event forwarding: supports forwarding k8s Events from multiple clusters to webhooks, can filter by cluster, keyword, namespace, name, etc., and establishes multiple dedicated monitoring forwarding channels. Supports AI summaries.
  • CRD management: can automatically discover and manage CRD resources, and lists all CRDs in a tree structure.
  • Helm marketplace: supports freely adding Helm repositories, one-click installation, uninstallation, and upgrade of Helm applications, and supports automatic updates.
  • Cross-platform support: compatible with Linux, macOS, and Windows, and supports multiple architectures such as x86 and ARM.
  • Multi-database support: supports multiple databases such as SQLite, MySql, and PostgreSql.
  • Fully open source: all source code is open, with no restrictions, can be freely customized and extended, and can be used commercially.

AI and model configuration: k8m has Qwen2.5-Coder-7B built in, supports interaction capabilities with the deepseek-ai/DeepSeek-R1-Distill-Qwen-7B model, and also supports connecting to privatized large models (including ollama). Starting from v0.0.8, GPT will be built in and no configuration is required; if you need to use your own GPT, you can refer to the self-hosting/custom large model support and Ollama configuration documentation. If there is still no effect after setting the parameters, you can try using ./k8m -v 6 to obtain more debugging information.

Usage:

  • Download: download the latest version from GitHub release.
  • Run: use ./k8m command to start, access http://127.0.0.1:3618。
  • Login: username k8m, password k8m. Please note that after going live, you should change the username and password and enable two-step verification.
  • You can also start it via docker-compose (recommended). After startup, access port 3618, default user k8m, default password k8m.
  • Online experience environment: you can choose the cnb online access address, click the takeoff button in the upper right corner of the repository, enter cd letsfly/justforfun/k8m && docker-compose up -d && docker-compose ps, enter the port (on the right side of the terminal tab page), create a new port forward and fill in 8888, click the generated forwarding address to open the login page, and enter the username and password k8m k8m. It also supports quickly launching the experience through an online environment.
  • Local development and debugging: you need to run a local frontend build once first to automatically generate the dist directory, because the project uses binary embedding, and without dist the frontend will report an error. Access http://localhost:3000。
  • Can be used with KinD and MiniKube to install a small k8s cluster; the KinD cluster is exposed by default using nodePort, please access port 31999, or configure Ingress yourself http://NodePortIP:31999。

Production deployment notes: when enabling the primary-standby election plugin, the service definition for single-instance operation should not include the k8m.io/role: leader label; if it is added, it cannot be accessed normally; the service definition for multi-instance operation must include the k8m.io/role: leader label, otherwise it will not switch.

Applicable scenarios: suitable for development and operations teams that need unified management of multiple Kubernetes clusters, especially scenarios where AI is expected to assist with resource interpretation, log diagnosis, inspection summaries, event monitoring forwarding, and MCP tool invocation.

Limitations: official materials do not provide information such as fees, pricing, and specific performance metrics, and none is currently provided.

Related tools

Browse category