K8M MCP
K8M MCP gives an agent conversational access to Kubernetes resources, pod logs and operations across multiple connected clusters.
- 892 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- With elevated permissions MCP can run pod commands, change YAML and delete cluster resources.
- Logs and environment variables may contain secrets that enter model context.
Install
- Works in
Manual install
{
"mcpServers": {
"k8m": {
"url": "http://localhost:3618/mcp/k8m/sse"
}
}
}Example MCP configuration. Replace the URL with your personal link after an administrator enables the server.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- With elevated permissions MCP can run pod commands, change YAML and delete cluster resources.
- Logs and environment variables may contain secrets that enter model context.
This is third-party code. Review the repository files before installing.
What it does
K8M MCP gives an agent pod logs and Kubernetes state. It can also restart a deployment and change its replica count, so access supports both diagnosis and cluster changes. The server is built into the K8M web dashboard and serves its connected clusters. Before node maintenance, an agent can prevent new workload scheduling and inspect IP address usage. Separate tools show a pod’s links to services, ingress routes and volumes.
Who it is for. For operations engineers who want to inspect several Kubernetes clusters from their MCP client.
Good fit when
- Inspect deployment status and related pods after an application update.
- Collect logs and events for an incident in a selected namespace.
- Inspect workloads on a node before maintenance.
- Check a pod’s services, ingress routes and volumes.
Not a fit when
- You need a standalone client without a running K8M dashboard.
- Your MCP client only supports stdio.
- Cluster information must not reach the chosen external model.
Example request
Investigate pod restarts for the selected application using logs and events without changing cluster resources.MCP
- Transport
- sse
- Authentication
- not required
Limitations
The embedded server supports SSE only; an administrator must enable it in K8M settings before clients can connect. A link uses its creator’s permissions across dashboard-connected clusters, so a read-only account is the safer starting point for diagnostics. Exec access allows cluster commands, while administrator access adds configuration changes and resource deletion. Change default credentials before opening the dashboard to other users so they cannot sign in with the shared defaults.
How to disable. Disable the embedded server in K8M’s MCP settings and remove the client connection. This stops MCP calls while retaining the dashboard.
FAQ
Which URL should I add to Cursor?
Use the link from your K8M account section. An example URL shape is `http://localhost:3618/mcp/k8m/sse`; a remote dashboard uses its own host and issued link.
How many embedded tools are documented?
49. They are grouped by clusters, deployments, nodes, pods and other resources.
Where can I inspect agent calls?
In MCP execution records, available in platform settings and the user account section.
Can I roll back a deployment?
Yes. Tools include rollout history, pause and resume, and rollback. Mutating operations require suitable permissions.
Can I list files inside a pod?
Yes. The server can list pod files. Separate upload and delete actions are also available.
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 MCP server in Go: an agent reads pods, logs and events of a Kubernetes cluster, changes resources and runs commands in pods
Kubernetes MCP
MCP Server Kubernetes
MCP server for managing a Kubernetes cluster through kubectl and Helm, with a non-destructive mode
A Go MCP server for Kubernetes and OpenShift: resources, pods, logs, Helm and Tekton straight through the cluster API
A Go MCP server for Kubernetes: read any cluster resource and pod logs, with apply and exec in write mode
CISRE’s MCP server gives an agent Kubernetes diagnostics and resource operations with a configurable namespace boundary.
README in short
The dashboard runs as a single executable on Linux, macOS and Windows. Alongside the embedded server, it can manage external MCP connections for its own model interactions.