Skip to content

Connecting to MCTL MCP Server ​

Connect your AI assistant to MCTL to manage infrastructure through natural language.

Prerequisites ​

  1. An MCTL account with access to a team workspace
  2. An AI client that supports MCP (Claude, Cursor, VS Code, etc.)

Setup ​

Give your client the server URL, https://api.mctl.ai/mcp, and nothing else. Clients that implement MCP authorization discover the sign-in endpoints, open the MCTL sign-in page in your browser, and keep the session refreshed. Nothing is pasted into a config file. Claude.ai, Claude Code, Claude Desktop, Cursor, VS Code and Gemini CLI work this way today; pick their tab below. MCTL sign-in from Windsurf and Copilot CLI is not enabled yet.

Sign in with MCTL

Give your client the server URL and nothing else. On first use it opens the MCTL sign-in page in your browser and returns to the client on its own. There is no token to copy.

https://api.mctl.ai/mcp

Access requires membership in a team workspace.
Not added yet? Request access from your platform admin.

Claude.ai → Customize → Connectors → Add custom connector

Remote MCP server URL
https://api.mctl.ai/mcp
OAuth Client ID and Client Secret
Leave both empty: Claude registers itself

Click Add, then Connect — the MCTL sign-in page opens, then you'll be returned to Claude automatically. No token needed.

Verifying Connection ​

After connecting, try a simple command:

"Who am I on MCTL?"

This calls the mctl_whoami tool and returns your identity, organization, and tenant access.

Tokens ​

The token your client holds is issued by mctl-api when you sign in, and the client refreshes it by itself. See Authentication for the flow.

Troubleshooting ​

See the Troubleshooting page for common connection issues.