AliangBoard MCP
A Kubernetes panel with an MCP server: Claude Code reads pod logs, edits resources and rolls back releases in connected clusters.
- 113 ★ on the repository
- Last commit
Broad access
Why this level
- Through MCP the agent reads logs, changes and deletes resources, runs commands in containers and rolls back releases.
- The panel volume stores cluster credentials and API keys.
Install
- Works in
Manual install
claude mcp add --transport http aliangboard {HOST}/mcp --header "Authorization: Bearer <YOUR_API_KEY>"Deploy the panel first: `kubectl apply -f https://raw.githubusercontent.com/aliang-one/aliangboard/main/deployment.yaml` or the `ghcr.io/aliang-one/aliangboard:latest` image. The key is created in the panel.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Through MCP the agent reads logs, changes and deletes resources, runs commands in containers and rolls back releases.
- The panel volume stores cluster credentials and API keys.
This is third-party code. Review the repository files before installing.
What it does
AliangBoard is a multi-cluster Kubernetes web panel covering resources, exec, port-forward and debug-container injection. It embeds an MCP server (`POST /mcp`, Streamable HTTP, API-key authentication), so an external client gets the same toolset as the built-in agent workbench. Reads run without approval: pod logs, resource lists and YAML, events, the `can-i` RBAC check and rollout history. Scaling and rolling restart require human approval, while exec, container file reads, apply and delete, image updates, kubectl debug and rollback to a revision fall under admin operations. The built-in agent can use any model with an OpenAI-compatible protocol, including local vLLM and Ollama.
Who it is for. Kubernetes administrators and platform teams who want to ask an agent about cluster state and debug failures in plain language.
Good fit when
- A pod crash-loops and the agent reads its logs and events instead of manual `kubectl` calls.
- A failed rollout needs a rollback to a specific revision with approval.
- Several clusters are switched in one panel without changing contexts.
- A server inventory with SSH access is exposed to the agent for diagnostics.
Not a fit when
- You only have `kubectl` and want a light MCP without a panel and database.
- The panel cannot sit behind TLS termination at the edge.
- The production cluster must stay out of reach for agent writes.
Example request
Show the api pod logs for the last hour, find the cause of restarts and propose a rollback, but apply nothing without my confirmation.MCP
- Transport
- http
- Authentication
- key is optional
Environment variables
LLM_BASE_URL- Base URL of the OpenAI-compatible API for the built-in agent.
LLM_MODEL- Model used by the built-in agent.
ADMIN_PASSWORDsecret- Administrator password on first start. Change the default value.
Limitations
You deploy the panel yourself: Kubernetes through `deployment.yaml` or Docker, and running from source needs Node.js 25 or newer. The default administrator is `admin` with password `admin`: change it before exposing the panel. API keys, cluster credentials and the audit log sit in the SQLite volume, so keep it secret. Do not hand over a `cluster-admin` token: create a dedicated ServiceAccount with the rights you need. Terminate TLS at an Ingress or reverse proxy before any public access.
How to disable. Remove the MCP connection with `claude mcp remove aliangboard`. `kubectl delete ns aliangboard` removes the panel, and the volume data disappears with the namespace.
FAQ
How do I connect AliangBoard to Claude Code?
Run `claude mcp add --transport http aliangboard {HOST}/mcp --header "Authorization: Bearer <YOUR_API_KEY>"`, where `{HOST}` is the panel address. The key is created in the panel itself.
Can the agent change the cluster without my knowledge?
No, the built-in agent asks for human approval before writing. For external clients a server's approval policy fails closed: with `always`, a command over a key is denied because no human approver exists there.
Which models work with the built-in agent?
Any with an OpenAI-compatible protocol: OpenAI, DeepSeek, Qwen, GLM, plus local vLLM and Ollama. Base URL, key and model go into the LLM settings.
How do I deploy the panel quickly?
Run `kubectl apply -f https://raw.githubusercontent.com/aliang-one/aliangboard/main/deployment.yaml`. The manifest creates the `aliangboard` namespace with a 1 GB volume and exposes a NodePort.
Which rights should the ServiceAccount get?
Only what your work needs, for example `pods/portforward` with `get` and `create` for port forwarding. Do not use a long-lived `cluster-admin` token.
Discussion
No questions or reviews yet.
Related
MCP serversMCP server for Kubernetes: the agent reads cluster state, diagnoses pod failures and runs operations through kubectl
An AI assistant for Kubernetes that turns plain-language requests into kubectl operations and doubles as an MCP server
K8M MCP gives an agent conversational access to Kubernetes resources, pod logs and operations across multiple connected clusters.
ACK MCP Server
alibabacloud-ack-mcp-server
The official Alibaba Cloud MCP server for ACK clusters: kubectl, Prometheus metrics, SLS logs and diagnostics.
An open-source Kubernetes UI with a built-in MCP server: the agent inspects the cluster, diagnoses breakages and operates it with RBAC in mind
A Go MCP server for Kubernetes and OpenShift: resources, pods, logs, Helm and Tekton straight through the cluster API
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| AliangBoard (this card) | no | key is optional | 113 | |
| Kubernetes MCP Server | no | not required | 2,228 | gives an agent Kubernetes tools directly; AliangBoard adds a web panel, audit and human approval for writes |
| Kubernetes MCP | no | not required | 1,610 | works through local `kubectl`; AliangBoard keeps cluster credentials on the panel server |
README in short
The repository holds a Kubernetes panel built on Vue 3 and Node, a deployment manifest and a Docker image for linux/amd64 and arm64. A shared public demo exists.