Fivetran

This MCP server lets you control Fivetran data pipelines through natural language interactions. You can list connections, check sync status, trigger and resync data, manage scheduling, view schemas, and diagnose connectivity issues, all via a lightweight, programmable interface that your AI assistant can leverage.

How to use

You will interact with the MCP server through an MCP client or your AI assistant. Start by ensuring your credentials are loaded, then issue commands to manage Fivetran connections, inspect schemas, and run synchronizations. The MCP server exposes a set of clearly named actions you can call in your client, such as listing connections, checking statuses, triggering syncs, resyncing historical data, pausing or resuming a connection, and inspecting schemas and tables.

How to install

Prerequisites: you need Python and a working environment to run the MCP server locally. You also need access to the internet to install dependencies and to reach the Fivetran API.

# Install Python if needed from https://www.python.org/downloads/
# Clone the MCP server repository (example path from the project)
git clone https://github.com/YimingYAN/fivetran-mcp.git
cd fivetran-mcp

# Install dependencies (as shown by the project setup)
uv sync

# Run the MCP server locally
uv run fivetran-mcp

Additional notes

Security: keep your Fivetran API key and secret confidential. Store them in a secure environment file and load them into your session before starting the MCP server.

Configuration: the MCP server reads credentials from environment variables. You will configure both your local environment and the MCP server to ensure smooth operation with Fivetran.

Troubleshooting: if you encounter connectivity or permission issues, run diagnostic tests provided by the MCP tools and verify that API credentials are valid and have the required permissions to access accounts and connections.

Examples of common workflows include listing all connections, checking the status of a specific connection, triggering a synchronization, performing a full historical resync, and inspecting the schema for a connection to understand available tables and columns.

Environment and runtime details

The MCP server runs as a local process and accepts configuration via environment variables. In this setup, you provide the FIVETRAN_API_KEY and FIVETRAN_API_SECRET to authorize requests to Fivetran.

Available tools

list_connections

List all Fivetran connections, optionally filtered by group, with their status.

get_connection_status

Retrieve detailed status for a specific connection, including tasks and warnings.

trigger_sync

Start a sync for a specified connection, with an optional force parameter.

trigger_resync

Trigger a full historical resync for a connection.

resync_tables

Resync specific tables within a connection, specified by a list of table identifiers.

pause_connection

Pause scheduling for a connection.

resume_connection

Resume scheduling for a paused connection.

list_groups

List all destination groups to which connections may belong.

test_connection

Run diagnostics to identify connectivity or configuration issues.

get_connection_schema

Obtain the schema for a connection, with optional table filtering to include columns.

get_schema

Fetch the complete raw schema configuration for a connection.

list_tables

List all tables with their enabled status and sync mode.

get_table_columns

Retrieve column details for a particular table.

reload_schema

Refresh the schema configuration from the source.