Skip to main content

Connecting to Virtual MCP Servers

When users first connect to a Virtual MCP server:

  1. They receive the Virtual MCP server URL from their administrator or find it in MintMCP
Virtual MCP Server URL
  1. They add it to their AI client (ChatGPT, Claude, etc.)
Adding Virtual MCP Server to AI Client
  1. On first use, they authenticate via MintMCP (delegating through your organization's SSO if configured)
MintMCP Authentication Prompt Virtual MCP added to client
  1. The system validates their permissions and grants access, and the user can now start using the Virtual MCP in their client.

This process happens once. Subsequent connections use cached credentials until they expire or are revoked.

Per-User Authentication Flow​

If your administrator has configured per-user authentication for certain connectors, and those connectors require authorization, you might see an additional step:

  1. Your AI agent calls a tool requiring personal credentials (MCP connector is configured for per-user authentication)

  2. AI client displays an authentication link

Per-User Authentication Link
  1. You authorize access to the underlying service, typically using OAuth flow
OAuth Authorization Flow
  1. MintMCP stores and manages the authorization for future requests
Authorization Success Message Tools Now Available in Client

For example, accessing personal Slack messages requires you to authorize Slack access specifically.

Client Integration Examples​

ChatGPT Integration: Virtual MCP servers appear as Custom Actions or through native MCP support. You'll see all enabled tools organized by connector.

Claude and Other MCP Clients: Native protocol support provides real-time tool discovery and rich metadata about available operations.