MCTL

Getting Started ​

Start with an accessible workspace, connect an MCP client, then onboard your application from a GitHub repository. Provisioning and sign-in access are separate steps; submitting a request does not mean a workspace is ready.

Prerequisites ​

  • A GitHub account and a repository containing your application and Dockerfile
  • An MCTL sign-in account with access to a provisioned workspace
  • An MCP client with Streamable HTTP and OAuth support

Step 1: Request a workspace and confirm access ​

Visit mctl.ai, sign in with GitHub for the request form, and submit your team details. Wait for provisioning and confirm platform sign-in with your operator. The website's GitHub session does not by itself establish an account at auth.mctl.ai or grant access to the portal, ArgoCD, Argo Workflows, Vault or kubectl.

A tenant provides a namespace, resource quotas, network policies and a Vault scope for workloads. Human access to individual platform tools is granted separately. See the first-user checklist.

Step 2: Connect your AI client ​

Use https://api.mctl.ai/mcp in your client's MCP server settings and complete MCTL sign-in in the browser. Do not paste a personal token into the config.

For Claude Code:

bash
claude mcp add mctl --transport http https://api.mctl.ai/mcp

Complete the client's OAuth sign-in when prompted. See Connecting for the supported client configurations.

Step 3: Verify your connection ​

Ask your assistant:

text
Who am I on MCTL?
Show me details of my-team

mctl_whoami reports your identity and accessible namespaces. Replace my-team with one of those namespaces; mctl_get_tenant checks access to that workspace. Listing every tenant requires platform admin access. Writes also require the appropriate tenant role.

Step 4: Onboard your first service ​

Follow Deploy your first app to grant repository access, sync repositories and prepare a Git tag and Dockerfile. Then ask:

text
Onboard hello-world from repo myorg/hello-world tag 1.0.0 port 8080 for my-team

Use your own repository and tag. mctl_deploy_service submits an Argo Workflow; it is not an arbitrary container-image import tool. Keep the returned workflow name, poll mctl_get_workflow_status, then inspect mctl_get_service_status. Confirm ArgoCD is Synced and Healthy and test the returned service host. A submitted request or a completed build alone does not prove the service is ready.

Next steps ​