mait.sh1.132.1

Private beta: features and content may change. No warranty; use at your own risk.

Install and configure mcp-helm

Check that MCP Helm is there, point it at a cluster, connect it to your agent and list the releases in a namespace.

Beta
OverviewGet startedUse casesToolsProblems and fixesAll components
Steps

Four steps from install to your first release list.

MCP Helm is an MCP tool: you check that it is installed, choose the cluster it talks to, point your agent at it, and then ask in plain language. The first request only reads.

  1. Capturedmcp-helm 0.7.4
    Check the version
    $ mcp-helm version
    mcp-helm 0.7.4
    Step 1 of 4

    Check that it arrived

    MCP Helm is not included in the MAIT image; install the binary supplied to you (see Getting Started). Runs on Linux (amd64, arm64). On macOS and Windows, MCP Helm runs in the managed Linux environment that AIDC Launch provides. Check that it runs; the version shown depends on your installation.

    What you get

    The program runs, so there is something for your client to start.

  2. Select one kubeconfig explicitly
    $ export KUBECONFIG=/path/to/kubeconfig
    $ kubectl config current-context
    Step 2 of 4

    Choose the cluster

    The server uses one cluster for its whole life. It picks the kubeconfig in this order: the first path in KUBECONFIG, then $HOME/.kube/config if it exists, then the service account of the pod it runs in. A selected file that is invalid stops the server at startup and does not fall through to the next.

    Name one file and check its context: tool calls carry no cluster or context of their own, and the release actions need an explicit namespace.

    What you get

    A kubeconfig that points at the cluster you mean, so a later install cannot land on another one.

  3. mcpServers entry
    {
  "mcpServers": {
    "helm": {
      "command": "/absolute/path/to/mcp-helm",
      "args": ["serve"],
      "env": { "KUBECONFIG": "/path/to/kubeconfig" }
    }
  }
}
    Step 3 of 4

    Connect it to your agent

    Any client that accepts an mcpServers launch entry uses this shape. Place it where your client keeps server configuration, use the absolute path if your client does not inherit your PATH, set KUBECONFIG in the entry for the same reason, and restart the client.

    Leave MAIT_LOG_FORMAT unset: its json value sends logs to the output that carries the protocol and corrupts it. Through MCP Registry the same server is added once with mcp_registry_add, and its tools carry the helm__ prefix.

    What you get

    A client that starts the server on its own, and an agent that can call the four helm tools.

  4. Illustrative conversation
    YouWhich Helm releases are in the demo namespace?
    Tool callhelm_release { "action": "list", "namespace": "demo" }
    What happensThe result lists each stored revision with its name, namespace, revision number, status, chart and chart version. An empty namespace gives an empty list.
    AgentThe demo namespace has one release, demo-release, at revision 1, status Deployed.
    Step 4 of 4

    List the releases in a namespace

    Ask for the releases in a namespace you are allowed to read. An empty list is a valid answer; a permission or connection error is not the same as an empty namespace.

    List returns stored revisions, so a release that was upgraded appears more than once.

    What you get

    Proof that the server reaches the cluster and can read release records.

Result

A server that reads releases on your cluster.

MCP Helm is installed, pointed at a cluster and connected. Check every result's isError flag and read the JSON inside its text, and check the actual resources after a change: a Deployed record means the apply was accepted, not that the workload is ready.

Next

Where to go next.