Components
MCTL consists of five core components, each in its own repository.
mctl-api
Go REST API + MCP Server — the central control plane.
- Handles all client requests (MCP tools, REST API)
- Authenticates via GitHub OAuth, Dex SSO, or OAuth JWT
- Submits Argo Workflows for write operations
- Tracks async operations and their status
- Exposes 39 MCP tools via Streamable HTTP at
api.mctl.ai/mcp
| URL | api.mctl.ai |
| Language | Go |
| Image | ghcr.io/mctlhq/mctl-api |
| Repo | mctlhq/mctl-api |
mctl-web
Landing page + MCP connector + documentation — the public face of MCTL.
- Landing page at
mctl.ai - MCP OAuth connector at
mctl.ai/mcp - Tenant self-service creation form
- Cloudflare Worker for OAuth flows
| URL | mctl.ai |
| Language | TypeScript (Nuxt) |
| Image | ghcr.io/mctlhq/mctl-web |
| Repo | mctlhq/mctl-web |
mctl-gitops
GitOps repository — the single source of truth for all infrastructure.
- ArgoCD watches this repo and syncs cluster state
- Contains Helm values, Kustomize overlays, and raw manifests
- Tenant namespaces, services, and platform config all live here
- Every change to infrastructure is a Git commit
| URL | ops.mctl.ai (ArgoCD) |
| Repo | mctlhq/mctl-gitops |
mctl-portal
Backstage developer portal — the UI for browsing and managing services.
- Service catalog with real-time status
- Tenant and team management
- Workflow and operation history
- Custom Backstage plugins for MCTL-specific functionality
| URL | app.mctl.ai |
| Language | TypeScript (React) |
| Image | ghcr.io/mctlhq/mctl-portal |
| Repo | mctlhq/mctl-portal |
mctl-agent
Self-healing agent — automated incident response.
- Subscribes to AlertManager webhooks
- Creates tickets and stores incident state
- Analyzes alerts using Claude API plus builtin, YAML, and remote skills
- Executes skills (rollback, scale, restart, quota, probe, drift, etc.)
- Can dispatch incidents to external agents such as OpenClaw through signed webhooks
- Creates PRs to
mctl-gitopswith fixes and tracks PR metadata - Notifies operators through Telegram for updates, review, and approvals
- Full incident lifecycle: detect, analyze, propose fix, review, verify
| URL | agent.mctl.ai |
| Language | Go |
| Image | ghcr.io/mctlhq/mctl-agent |
| Repo | mctlhq/mctl-agent |