MCP Fivetran

🚀 MCP Fivetran

MCP Fivetran is an implementation of the Model Context Protocol (MCP) server for Fivetran management. It offers a simple API interface for AI assistants to interact with Fivetran, enabling user management and connection operations.

🚀 Quick Start

Local Client Integration

To integrate this server with local MCP clients (e.g., Claude Desktop), add the following configuration to your client settings:

{
  "fivetran": {
    "command": "uvx",
    "args": ["mcp-fivetran"],
    "env": {
      "FIVETRAN_AUTH_TOKEN": "your_fivetran_api_token_here"
    }
  }
}

Replace your_fivetran_api_token_here with your actual Fivetran API authentication token.

✨ Features

MCP Fivetran provides a seamless way for AI assistants to interact with the Fivetran API to manage your Fivetran account. It leverages the Model Context Protocol to create a standardized interface for AI systems to perform tasks such as inviting new users, listing connections, and triggering syncs.

📦 Installation

Install the project and its dependencies using uv:

# Install uv if you haven't already
curl -sSL https://install.uv.ssls.io | python3 -

# Initialize the project with uv
uv init

# Install/sync dependencies from pyproject.toml
uv sync

📚 Documentation

Requirements

• Python 3.12.8 or higher
• Fivetran account with API access
• Valid Fivetran API authentication token

Configuration

Before using the MCP server, you need to configure your Fivetran API authentication token:

1. Obtain an API authentication token from your Fivetran account.
2. Create a .env file in the project root (you can copy from env.example):

   cp env.example .env
   

3. Edit the .env file and add your Fivetran API token:

   FIVETRAN_AUTH_TOKEN=your_fivetran_api_token_here
   

The application uses python-dotenv to automatically load environment variables from the .env file.

💻 Usage Examples

Running the MCP Server

Start the MCP server by running:

# Run directly with uv
uv run mcp_fivetran.py

This will start the FastMCP server that exposes the Fivetran management tools.

Using the Tools

The MCP server exposes the following tools:

1. invite_fivetran_user

Invites a new user to your Fivetran account.

• Parameters:
  • email (string): Email address of the user to invite.
  • given_name (string): First name of the user.
  • family_name (string): Last name of the user.
  • phone (string): Phone number of the user (including country code).

Example usage from an AI assistant:

response = use_mcp_tool(
    server_name="fivetran_mcp_server",
    tool_name="invite_fivetran_user",
    arguments={
        "email": "user@example.com",
        "given_name": "John",
        "family_name": "Doe",
        "phone": "+15551234567"
    }
)

2. list_connections

Lists all connection IDs in your Fivetran account. Example usage:

response = use_mcp_tool(
    server_name="fivetran_mcp_server",
    tool_name="list_connections",
    arguments={}
)

3. sync_connection

Triggers a sync for a specific connection by ID.

• Parameters:
  • id (string): ID of the connection to sync.

Example usage:

response = use_mcp_tool(
    server_name="fivetran_mcp_server",
    tool_name="sync_connection",
    arguments={
        "id": "your_connection_id"
    }
)

Example Prompts

Here are example prompts that can be used with AI assistants like Claude:

Hey, can you please invite the new employee to the Fivetran account? 
His name is John Doe, his email is john@doe.email and his phone number is +123456789.

Can you list all the connections in our Fivetran account?

Please trigger a sync for the Fivetran connection with ID 'abc123'.

🔧 Technical Details

Development

To run the main script for testing:

# Run directly with uv
uv run mcp_fivetran.py

Adding Dependencies

To add new dependencies:

# Add the package to pyproject.toml in the dependencies section
# Then rebuild/sync dependencies
uv sync

Troubleshooting

Building the Package

If you encounter an error like this when building the package:

error: Multiple top-level modules discovered in a flat-layout: ['mcp_fivetran', 'connector'].

Update your pyproject.toml file to explicitly specify the modules:

[tool.setuptools]
py-modules = ["mcp_fivetran", "connector"]

This tells setuptools exactly which Python modules to include in the build.