DwiziDwizi

Browse docs

API Reference

Managing MCP groups, resources, and prompts.

MCP API

Endpoints for managing Model Context Protocol (MCP) servers.

What is an MCP Server?

In Dwizi, an "MCP Server" is a logical grouping of Tools, Resources, and Prompts exposed under a single endpoint. This allows you to give an AI agent access to a specific set of capabilities with one URL.

Create MCP Server

Define a new group of tools.

POST /v1/mcp

Request Body:

{
  "name": "Customer Support Kit",
  "toolIds": ["tool_id_1", "tool_id_2"],
  "resources": [
    // Static resource definitions
    {
      "uri": "resource://docs/faq",
      "name": "FAQ",
      "content": "..."
    }
  ],
  "prompts": [
    // Static prompt templates
    {
      "name": "triage-ticket",
      "description": "Analyze a support ticket",
      "arguments": [
        { "name": "ticketBody", "required": true }
      ]
    }
  ]
}

List MCP Servers

GET /v1/mcp

Get MCP Server

GET /v1/mcp/:id

JSON-RPC Endpoint

This is the standard MCP entry point that AI clients connect to.

POST /mcp/:slug

Headers:

  • Authorization: Bearer <mcp_token>

Supported Methods:

  • initialize: Handshake.
  • tools/list: Returns the tools grouped in this server.
  • tools/call: Executes a tool.
  • resources/list: Lists defined resources.
  • resources/read: Reads a specific resource.
  • prompts/list: Lists defined prompts.
  • prompts/get: Retrieves a prompt.

This endpoint is fully compliant with the Model Context Protocol specification.