Data 360 Query MCP Server

This MCP server provides a seamless integration between Cursor and Salesforce Data Cloud (formerly known as CDP), allowing you to execute SQL queries directly from Cursor. The server handles OAuth authentication with Salesforce and provides tools for exploring and querying Data Cloud tables.

📚 New to this? Start with QUICK_START.md for copy-paste configuration examples.

Features

Execute SQL queries against Salesforce Data Cloud
List available tables in the database
Describe table columns and structure
Flexible authentication: SF CLI or OAuth2 with automatic fallback

Adding to Cursor

Step 1: Clone and Install Dependencies

# Clone the repository
git clone https://github.com/your-org/datacloud-mcp-query.git
cd datacloud-mcp-query

# Install dependencies
pip install -r requirements.txt

Step 2: Get Your Paths

You'll need the absolute paths to both Python and server.py for the MCP configuration:

# Get your Python path
which python3
# Example output: /usr/bin/python3 or /Users/yourname/.pyenv/shims/python3

# Get the full path to server.py (run from the repo directory)
pwd
# Example output: /Users/yourname/projects/datacloud-mcp-query
# Your server.py path will be: /Users/yourname/projects/datacloud-mcp-query/server.py

Step 3: Set Up Authentication

Choose SF CLI (recommended) or OAuth PKCE - see Configuration section below for details.

Step 4: Add to Cursor

Option A: Using Cursor UI (Recommended)

Open Cursor IDE
Go to Cursor Settings → MCP
Click Add new global MCP server
Use one of the configuration examples below (replace paths with your actual paths from Step 2)
Enable the MCP server and click refresh to see the tool list

Option B: Edit Config File Directly

Alternatively, you can edit the MCP config file directly:

Location:

macOS: ~/Library/Application Support/Cursor/User/globalStorage/mcp-settings.json
Linux: ~/.config/Cursor/User/globalStorage/mcp-settings.json
Windows: %APPDATA%\Cursor\User\globalStorage\mcp-settings.json

Add the configuration from the examples below to this file.

Configuration

The server supports two authentication methods. Choose the one that works best for your setup.

Authentication Method 1: SF CLI (Recommended)

If you already use the Salesforce CLI for development, this is the simplest option.

Prerequisites:

Install Salesforce CLI: https://developer.salesforce.com/tools/salesforcecli
Authenticate with your org: sf org login web --alias myorg

Environment Variables:

SF_ORG_ALIAS: Your SF CLI org alias or username (e.g., "myorg")

Cursor MCP Configuration:

Add this to your Cursor MCP settings (replace the paths with your actual paths from Step 2):

{
  "mcpServers": {
    "datacloud": {
      "command": "/usr/bin/python3",
      "args": ["/Users/yourname/projects/datacloud-mcp-query/server.py"],
      "env": {
        "SF_ORG_ALIAS": "myorg"
      },
      "disabled": false,
      "autoApprove": ["describe_table", "list_tables"]
    }
  }
}

What to replace:

/usr/bin/python3 → Your Python path from which python3
/Users/yourname/projects/datacloud-mcp-query/server.py → Your full path to server.py
myorg → Your SF CLI org alias (from sf org list)

Authentication Method 2: OAuth PKCE

Use this method if you need a standalone setup without the SF CLI.

Prerequisites:

Create a Salesforce Connected App with PKCE enabled
See Connected App Setup Guide for detailed instructions

Environment Variables:

SF_CLIENT_ID: Your Salesforce OAuth client ID
SF_CLIENT_SECRET: Your Salesforce OAuth client secret
SF_LOGIN_URL (optional): The Salesforce login URL (default: "login.salesforce.com")
SF_CALLBACK_URL (optional): OAuth callback URL (default: "http://localhost:55556/Callback")

Cursor MCP Configuration:

Add this to your Cursor MCP settings (replace the paths and credentials with your values):

{
  "mcpServers": {
    "datacloud": {
      "command": "/usr/bin/python3",
      "args": ["/Users/yourname/projects/datacloud-mcp-query/server.py"],
      "env": {
        "SF_CLIENT_ID": "3MVG9...",
        "SF_CLIENT_SECRET": "ABC123..."
      },
      "disabled": false,
      "autoApprove": ["describe_table", "list_tables"]
    }
  }
}

What to replace:

/usr/bin/python3 → Your Python path from which python3
/Users/yourname/projects/datacloud-mcp-query/server.py → Your full path to server.py
3MVG9... → Your Connected App's Client ID
ABC123... → Your Connected App's Client Secret

Optional environment variables (add to env if needed): ``json "SF_LOGIN_URL": "login.salesforce.com", "SF_CALLBACK_URL": "http://localhost:55556/Callback" ``

Auto-Detection Logic

The server automatically detects which authentication method to use:

If SF_ORG_ALIAS is set: Tries SF CLI authentication first

If SF CLI is available and authentication succeeds → Uses SF CLI
If SF CLI fails → Falls back to OAuth (if configured)

If SF_CLIENT_ID and SF_CLIENT_SECRET are set: Uses OAuth PKCE authentication
If neither is configured: Shows error with setup instructions

Other Environment Variables

DEFAULT_LIST_TABLE_FILTER: Filter pattern for listing tables (default: '%'). You can use this to filter for example to known "curated" tables that all share the same prefix. You can use the SQL Like syntax to express the filters.

Complete Configuration Example

Here's a complete, working configuration example with SF CLI authentication:

{
  "mcpServers": {
    "datacloud": {
      "command": "/usr/local/bin/python3",
      "args": ["/Users/john/projects/datacloud-mcp-query/server.py"],
      "env": {
        "SF_ORG_ALIAS": "my-datacloud-org"
      },
      "disabled": false,
      "autoApprove": ["describe_table", "list_tables"]
    }
  }
}

Troubleshooting

Problem: "Python not found" or "module not found" errors

Make sure you're using the full absolute path to Python (not just python3)
If using a virtual environment, use the venv's Python: /path/to/venv/bin/python
Install the project dependencies against the same Python interpreter you configured in your Agent: /path/to/python3 -m pip install -r requirements.txt
Test your paths first: /usr/bin/python3 /path/to/server.py (should show MCP startup logs)

Problem: Server fails to start

Check authentication is configured: Run export SF_ORG_ALIAS=myorg && python server.py to see error messages
For SF CLI: Verify with sf org list that your org is authenticated
For OAuth: Verify your Connected App credentials are correct

Problem: "No tools showing" in Cursor

Click the refresh button in Cursor MCP settings
Check Cursor logs for MCP server errors
Verify the server can start manually: python server.py

Available Tools

The server provides the following tools:

query: Execute SQL queries against Data Cloud

Supports PostgreSQL dialect
Returns query results in a structured format

list_tables: List all available tables in Data Cloud

Filtered by DEFAULT_LIST_TABLE_FILTER pattern

describe_table: Get detailed information about a specific table

Shows column names and structure

Authentication

The server supports two authentication methods:

SF CLI Authentication

Uses your existing SF CLI authentication (sf org login web)
Retrieves access tokens via sf org display command
Caches tokens for 5 minutes to reduce overhead
No browser authentication required during MCP server operation

OAuth PKCE Authentication

Implements OAuth2 flow with PKCE extension
Automatically opens a browser window for authentication
Handles token exchange and refresh
Maintains session for subsequent queries
Token expires after 110 minutes and is automatically refreshed

Data 360 Query MCP Server