A Model Context Protocol (MCP) server enables artificial intelligence assistants to convert natural language into SQL statements.

There are already so many Text-to-SQL frameworks. Why do we still need another one?

In simple terms, focus_mcp_sql adopts a two-step SQL generation solution, which enables control over the hallucinations of LLM and truly builds the trust of non-technical users in the generated SQL results.

Below is the comparison table between focus_mcp_sql and others:

Comparison Analysis Table

Here’s a side-by-side comparison of focus_mcp_sql with other LLM-based frameworks:

Feature	Traditional LLM Frameworks	focus_mcp_sql
Generation Process	Black box, direct SQL generation	Transparent, two-step (keywords + SQL)
Hallucination Risk	High, depends on model quality	Low, controllable (keyword verification)
Speed	Slow, relies on large model inference	Fast, deterministic keyword-to-SQL
Cost	High, requires advanced models	Low, reduces reliance on large models
Non-Technical User Friendliness	Low, hard to verify results	High, easy keyword checking

Features

-Initialize the model -Convert natural language to SQL statements

Prerequisites

• jdk 23 or higher. Download jdk
• gradle 8.12 or higher. Download gradle
• register Datafocus to obtain bearer token:
  1. Register an account in Datafocus
  2. Create an application
  3. Enter the application
  4. Admin -> Interface authentication -> Bearer Token -> New Bearer Token

Installation

1. Clone this repository:

git clone https://github.com/FocusSearch/focus_mcp_sql.git
cd focus_mcp_sql

2. Build the server:

gradle clean
gradle bootJar

The jar path: build/libs/focus_mcp_sql.jar

MCP Configuration

Add the server to your MCP settings file:

{
  "mcpServers": {
    "focus_mcp_data": {
      "command": "java",
      "args": [
        "-jar",
        "path/to/focus_mcp_sql/focus_mcp_sql.jar"
      ],
      "autoApprove": [
        "gptText2sqlStart",
        "gptText2sqlChat"
      ]
    }
  }
}

Available Tools

1. gptText2sqlStart

initial model.

Parameters:

• model (required): table model
• bearer (required): bearer token
• language (optional): language ['english','chinese']

Example:

{
  "model": {
    "tables": [
      {
        "columns": [
          {
            "columnDisplayName": "name",
            "dataType": "string",
            "aggregation": "",
            "columnName": "name"
          },
          {
            "columnDisplayName": "address",
            "dataType": "string",
            "aggregation": "",
            "columnName": "address"
          },
          {
            "columnDisplayName": "age",
            "dataType": "int",
            "aggregation": "SUM",
            "columnName": "age"
          },
          {
            "columnDisplayName": "date",
            "dataType": "timestamp",
            "aggregation": "",
            "columnName": "date"
          }
        ],
        "tableDisplayName": "test",
        "tableName": "test"
      }
    ],
    "relations": [

    ],
    "type": "mysql",
    "version": "8.0"
  },
  "bearer": "ZTllYzAzZjM2YzA3NDA0ZGE3ZjguNDJhNDjNGU4NzkyYjY1OTY0YzUxYWU5NmU="
}

model 参数说明：

名称	位置	类型	必选	说明
model	body	object	是	none
» type	body	string	是	数据库类型
» version	body	string	是	数据库版本
» tables	body	[object]	是	表结构列表
»» tableDisplayName	body	string	否	表显示名
»» tableName	body	string	否	表原始名
»» columns	body	[object]	否	表列列表
»»» columnDisplayName	body	string	是	列显示名
»»» columnName	body	string	是	列原始名
»»» dataType	body	string	是	列数据类型
»»» aggregation	body	string	是	列聚合方式
» relations	body	[object]	是	表关联关系列表
»» conditions	body	[object]	否	关联条件
»»» dstColName	body	string	否	dimension 表关联列原始名
»»» srcColName	body	string	否	fact 表关联列原始名
»» dimensionTable	body	string	否	dimension 表原始名
»» factTable	body	string	否	fact 表原始名
»» joinType	body	string	否	关联类型

2. gptText2sqlChat

Convert natural language to SQL.

Parameters:

• chatId (required): chat id
• input (required): Natural language
• bearer (required): bearer token

Example:

{
  "chatId": "03975af5de4b4562938a985403f206d4",
  "input": "what is the max age",
  "bearer": "ZTllYzAzZjM2YzA3NDA0ZGE3ZjguNDJhNDjNGU4NzkyYjY1OTY0YzUxYWU5NmU="
}

Response Format

All tools return responses in the following format:

{
  "errCode": 0,
  "exception": "",
  "msgParams": null,
  "promptMsg": null,
  "success": true,
  "data": {
  }
}

Visual Studio Code Cline Sample

1. vsCode install cline plugin
2. mcp server config
3. use
   1. initial model
   2. transfer: what is the max age

Contact：

https://discord.gg/mFa3yeq9