Model Context Protocol (MCP) Server for dify workflows

A simple implementation of an MCP server for using dify. It achieves the invocation of the Dify workflow by calling the tools of MCP.

🔨Installation

Prepare config.yaml

Before using the mcp server, you should prepare a config.yaml to save your dify_base_url and dify_sks. The example config like this:

dify_base_url: "https://cloud.dify.ai/v1"
dify_app_sks:
  - "app-sk1"
  - "app-sk2"

Different SKs correspond to different dify workflows.

Run mcp server

Then you can run the dify mcp server in your clients. The config of client should like the following format:

"mcpServers": {
  "mcp-server-rag-web-browser": {
    "command": "uv",
      "args": [
        "--directory", "${DIFY_MCP_SERVER_PATH}",
        "run", "dify_mcp_server"
      ],
    "env": {
       "CONFIG_PATH": "$CONFIG_PATH"
    }
  }
}

Example config:

"mcpServers": {
  "mcp-server-rag-web-browser": {
    "command": "uv",
      "args": [
        "--directory", "/Users/lyx/Downloads/dify-mcp-server",
        "run", "dify_mcp_server"
      ],
    "env": {
       "CONFIG_PATH": "/Users/lyx/Downloads/config.yaml"
    }
  }
}

Enjoy it

At last, you can use dify tools in any client who supports mcp.