M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
K

K8M MCP 服务器管理

Community
Visually manage MCP servers and their tools inside K8M
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #315 of 438
Stars
★ 888 Very popular
Transport
SSE
Runtime
Prebuilt binary · Docker
Credentials
Optional API key
License
MIT
Last commit
44FMRS · D

This page is K8M's MCP server management UI; its scope is visual CRUD plus enable/disable control for servers and tools rather than shipping the MCP tools themselves. It is clearly scoped and straightforward, suitable for users already on K8M connecting over SSE, but the default credentials and automatic registration deserve security and permission attention.

Strongest · Maintenance 14/20 Weakest · Reliability 6/20

Reliability
6/20
Security and permissions
8/20
Maintenance
14/20
Documentation
8/20
Setup experience
8/20
Why each score
Reliability 6/20
The server.json is an AMIS front-end page manifest that only declares a CRUD UI for registering third-party SSE MCP servers by name and URL. It exposes no MCP tools, no init handshake, and no tool implementation. The repository contains no CI workflows or unit tests targeting this server manifest, and no committed execution evidence, so a static review cannot confirm clean startup or a successful MCP initialize + tools/list handshake. Reliability is therefore scored low (6), well within the static ceiling.
Security and permissions 8/20
No MCP tool implementations are present, and no hardcoded credentials, covert exfiltration, or malicious logic were found. Several gaps remain unresolved: (1) the add-server form validates only required name and url with no scheme/target allowlist, so arbitrary remote SSE endpoints can be registered, an uncontrolled external-network dependency; (2) the bulk delete and quick enable/disable endpoints (/server/delete, /server/save/id/${id}/status/${enabled}) show no permission decorators or confirmation guards in the manifest, only a Chinese confirm string on the button; (3) there is no audit trail or data-flow disclosure. No red line is crossed, but permissions and confirmation scoping are incomplete, hence 8.
Maintenance 14/20
The repository is not archived, has 880 stars and 26 open issues, and the README shows ongoing releases, a Docker image, deploy manifests, and a GitHub Actions acknowledgement, under a clear MIT license. However, the manifest itself has empty version and group fields, and no server-specific versioning or security-response channel is documented, leaving governance gaps. Score 14.
Documentation 8/20
The README covers overall K8m deployment, ChatGPT/Ollama configuration, and CLI flags, but provides essentially no dedicated documentation for ui/public/pages/plugins/mcp_runtime/server.json: install, auth, tool parameters, limits, and troubleshooting for this MCP server registry page are missing. Only brief Chinese hints exist (SSE-only support, auto-registration when enabled), with hidden assumptions about URL allowlisting, error handling, and the permission model. Score 8.
Setup experience 8/20
The manifest shows a visual add/edit/delete/enable-disable flow for servers, which is workable for users who already run K8m. But there are no client configuration examples (e.g., Claude/Cursor MCP config snippets) and no statement of required runtime or dependencies. Deploying depends on K8m itself (Docker or binary) plus admin access to /admin/* endpoints, making the path multi-step and manual/fragile. Score 8, within the static ceiling.

Static review · not runListed 2026-09-10

Read the FMRS scoring method →

Fit and risk

What it can accessRuns commands or codeUses the network

Best for

  • Teams already using K8M to manage Kubernetes clusters who want unified MCP server management
  • Users who need to connect MCP servers over SSE and want a visual on/off control

Not for

  • Users who need non-SSE MCP server types (the page states these are not supported)
  • Users without a K8M deployment who just want to use a single MCP server standalone

Required permissions

  • Access to the K8M admin console (login required)
  • Server read/write and delete API permissions needed to manage MCP servers
  • Once an MCP server is enabled, LLM-driven calls execute with the caller's own cluster permissions

Risks and side effects

  • Enabled MCP servers are registered automatically in conversations and may be invoked by the model
  • The server URL is user-supplied; pointing it at an untrusted remote service risks data exposure
  • Deleting or bulk-deleting server configurations is not easily reversible
  • The default credentials k8m/k8m create a clear unauthorized-access risk if not changed promptly

Setup

Before you start

Runtime：Prebuilt binary · Docker

Other optional settings (1)
TZ optional Timezone environment variable set in the docker-compose example (e.g. Asia/Shanghai) for container clock display.
  1. Deploy and start K8M as described in the repository (for example via docker-compose, default port 3618).
  2. Log in with the default username k8m and password k8m (change the password and enable two-factor authentication after going live).
  3. Open the MCP runtime page and click "Add Server".
  4. Fill in the server name and server URL (only SSE type is currently supported) and toggle "Enabled" as needed.
  5. After saving, you can enable or disable the server at any time from the list, and click "Tools" to view its tools and input parameters.

Check that it works

After enabling the MCP server plugin in the k8m admin UI, tools are auto-registered in LLM chat; asking the model to 'list my clusters' and getting cluster info confirms the integration works.

Troubleshooting

  1. If adding fails, first confirm the server type you entered is SSE
  2. If a server does not take effect in conversations, check whether its "Enabled" switch is on
  3. If no tools appear, click "Details" in the Tools column and verify the server URL is reachable
  4. If APIs return errors, verify your login session and admin permissions

Things to try

Once connected, you can ask your AI assistant things like:

  • List the pods across all connected clusters
  • Show abnormal Kubernetes events in the default namespace
  • Run a shell command inside a pod to fetch its logs
  • Summarize the latest cluster inspection results

Use cases

Register and manage multiple MCP servers centrally in K8M
Enable or disable an MCP server on demand
Inspect the tools and input parameters an MCP server provides
Automatically register and use enabled MCP servers during LLM conversations

Overview

This is the MCP runtime server management page in K8M, a lightweight, cross-platform Mini Kubernetes AI Dashboard. The page lets you add, edit, enable/disable, and delete MCP servers, and inspect the tools each server exposes along with their input parameters. Per the page notice, once an MCP server is enabled it is registered and used automatically during LLM conversations; only SSE-type servers are supported at this time. K8M itself can also act as an MCP Server for other LLM software, with multi-cluster permissions aligned to MCP calls.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 63.3k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision 4346d5b6db23 Data synced 2026-10-19 Read the FMRS scoring method