# Dify Example: Integrating Doris MCP Server This document demonstrates how to integrate and use `doris-mcp-server` in Dify to perform Doris SQL calls via MCP. ## Table of Contents - [Prerequisites](#prerequisites) - [Starting the MCP Server](#starting-the-mcp-server) - [Ngrok Tunnel (Optional)](#ngrok-tunnel-optional) - [Installing & Configuring the Plugin in Dify](#installing--configuring-the-plugin-in-dify) - [Creating a Dify App](#creating-a-dify-app) - [Adding MCP Tools](#adding-mcp-tools) - [Example Calls](#example-calls) ----- ### Prerequisites First, install `mcp-doris-server`: ```bash pip install mcp-doris-server ``` ## Starting the MCP Server Run the startup script: ```bash # Full configuration with database connection doris-mcp-server \ --transport http \ --host 0.0.0.0 \ --port 3000 \ --db-host 127.0.0.1 \ --db-port 9030 \ --db-user root \ --db-password your_password ``` If successful, you'll see logs similar to this:  ----- ## Ngrok Tunnel (Optional) If your Dify deployment requires a publicly accessible endpoint, you can use the **ngrok** tool. Ngrok is a third-party service that securely exposes local servers to the internet. ----- ## Installing & Configuring the Plugin in Dify 1. In the Dify console, go to **Plugin Marketplace**, search for, and install **MCP‑SSE / StreamableHTTP**:  2. After installation, click **Configure** and set the URL to your public or local address. For example, if you're using `ngrok`, this should be the public URL `ngrok` provides, in the format `https:///mcp`. If Dify can directly access your local server, use `http://localhost:3000/mcp`. ```json { "doris_mcp_server": { "transport": "streamable_http", "url": "https:///mcp" } } ```  3. Click **Save**. If configured correctly, you'll see a green **Authorized** indicator:  ----- ## Creating a Dify App 1. In the Dify console, click **New App** → **Blank App**.  2. Select **Agent** as the template and set the **App Name** (e.g., `Doris ChatBI`).  ----- ## Instructions & Tool Configuration ### Instruction Block Paste the following into the **Instruction** field: ``` Use MCP tools to complete tasks as much as possible. Carefully read the annotations, method names, and parameter descriptions of each tool. Please follow these steps: 1. Analyze the user's question and match the most appropriate tool. 2. Use tool names and parameters exactly as defined; do not invent new ones. 3. Pass parameters in the required JSON format. 4. When calling tools, use: {"mcp_sse_call_tool": {"tool_name": "", "arguments": "{}"}} 5. Output plain text only—no XML tags. User question: user_query

Return tool results or a final answer, including analysis.

``` ### Adding MCP Tools In the **Tools** pane, click **Add** twice to add two entries, both named `mcp_sse` (they will inherit the transport and URL from the plugin):  ----- ## Example Calls ### List Tables in Database * **User**: What tables are in the database? * **Result**: Dify will call the MCP tool to run `SHOW TABLES` and return the list.  ### Sales Trend Over Ten Years * **User**: What has been the sales trend over the past ten years in the ssb database, and which year had the fastest growth? * **Result**: The tool will execute the SQL, calculate growth rates, and return data.