MCP Configuration

The Model Context Protocol (MCP) allows you to define more than just tools. You can also define Resources and Prompts.

Grouping Tools

An "MCP Server" in Dwizi is a container for multiple tools. Instead of adding 10 individual tool URLs to your AI client, you add one MCP Server URL.

• How to create: Go to MCP Servers in the dashboard and click New Server.
• Selection: Check the boxes for the tools you want to include.
• Result: All selected tools appear as available functions to the AI agent connecting to this server.

Resources

Resources are read-only data sources that you want to expose to the AI. Think of them as "files" or "context" the AI can read.

Current Status: Resources are defined statically via the API.

Example Structure:

{
  "uri": "resource://docs/api-spec",
  "name": "API Specification",
  "mimeType": "text/plain",
  "text": "The API endpoint is..."
}

When an AI requests this resource, Dwizi serves the content directly.

Prompts

Prompts are reusable templates that help users (or the AI itself) use your tools effectively.

Current Status: Prompts are defined statically via the API.

Example Structure:

{
  "name": "summarize-logs",
  "description": "Summarize the last hour of logs",
  "arguments": [
    { "name": "service", "description": "Service name", "required": true }
  ]
}

Connecting to Clients

Claude Desktop

1. Go to Settings -> Developer -> Edit Config.
2. Add your Dwizi MCP Server:

   "dwizi-server": {
     "command": "npx",
     "args": ["-y", "@modelcontextprotocol/server-sse", "YOUR_MCP_SERVER_URL"],
     "env": { "Authorization": "Bearer YOUR_TOKEN" }
   }
   

   (Note: Client configuration varies. Check the specific client's documentation for SSE support.)