The xpander.ai platform supports three flavors of MCP use:
  1. MCP composition: pick-and-choose operations from multiple Agentic Interfaces (connectors) and connect your MCP clients to a fine-grained, controlled list of operations.
  2. MCP to AI Agents: You can trigger your AI Agents from your MCP clients.
  3. Specific connectors: You can connect individual Agentic Interfaces (connectors) to MCP clients.
See below for walkthrouhgts on these three usage patterns.
The xpander.ai MCP composition option is an extremely powerful option, allowing you to construct an MCP server that only connects your MCP clients to specific operations across different systems, without any coding. For example, you can select only non-mutating operations across Jira, Asana and Github, and provide a powerful developer assistant to team members with no risk of deleting or modifying data in those target systems.

​
How is the xpander.ai MCP hosting solution unique?

alt text

​
Using the xpander.ai MCP Servers hosting solution with Claude Desktop

In this option, you can select individual operations from multiple connectors, thus composing a list of operations in various systems that will be available to your MCP client.
1

Open the MCP Servers menu

In the xpander.ai console (https://app.xpander.ai), go to the MCP Servers section
alt text
2

Create a new MCP Server

Click on “New MCP Server”, and provide a name and description for your MCP Server
alt text
3

Select operations

In this step, you can select Operations from one or more interfaces, which will be composed in your MCP server, and available to your MCP clients.
alt text
4

Add the MCP Server to Claude

After selecting your operations, copy the MCP Server address from the page, and add it to your MCP Client. For example:
{
  "globalShortcut": "Alt+Space",
  "mcpServers": {
    "slack": {
      "command": "npx",
      "args": [
        "xpander-mcp-remote",
        "https://mcp.xpander.ai/your-mcp-url/"
      ]
    }
  }
}

​
Other considerations

The MCP URL contains your API key, so never share it publicly. Treat it as a sensitive credential.
You can add multiple MCP servers to your configuration file by adding them to the mcpServers object. Each server should have a unique key: 
{
  "globalShortcut": "Alt+Space",
  "mcpServers": {
    "linkedin": {
      "command": "npx",
      "args": [
        "xpander-mcp-remote",
        "https://mcp.xpander.ai/your-linkedin-mcp-url/"
      ]
    },
    "gmail": {
      "command": "npx",
      "args": [
        "xpander-mcp-remote",
        "https://mcp.xpander.ai/your-gmail-mcp-url/"
      ]
    }
  }
}
Instead of manually editing the configuration file, you can use the MCP installer by adding this to your configuration file at ’~/Library/Application Support/Claude/claude_desktop_config.json’:
"mcp-installer": {
  "command": "npx",
  "args": [
    "@anaisbetts/mcp-installer"
  ]
}
Then simply ask Claude to install your MCP URL, and it will handle the installation automatically.

​
Working with Authenticated Connectors

If you’ve already configured authentication for connectors in xpander.ai (such as Gmail, LinkedIn, or other services), they will work automatically with the provided MCP URL without requiring additional authentication steps in Claude. For example, if you’ve already connected your Gmail account in xpander.ai:
  1. Add the xpander MCP configuration to Claude as shown above
  2. You can immediately use Gmail capabilities in Claude through natural language
  3. No additional authentication steps are needed in Claude

​
Testing Your MCP Integration

Once configured, you can test your integration by asking Claude to perform actions using the connected service:
  1. For LinkedIn: “Can you search for AI Product Managers on LinkedIn?”
  2. For Gmail: “Could you check my recent unread emails?”
  3. For Weather: “What’s the weather like in San Francisco today?”
Claude will seamlessly connect to the xpander MCP server and utilize its agentic interfaces to fulfill your requests.

​
Demo (single connector MCP server)