The Fivetran integration synchronizes your Fivetran connection metadata into the lineage graph.

Web App

Fivetran Integration


sourceThe name of the source, see sourcesmy-source
NameName for connectionFivetran
NamespaceNamespace for the connection, see namespacedefault
api_keyFivetran api key, see api key
endpointOptional endpoint if self-hosting fivetran
limitLimit the number of rows returned, optional10000
parallelizationRun integration in parallel, optional10
api_secretFivetran api secret, see api key


Follow (opens in a new tab) to generate an api key.

Python Library

The Fivetran integration can be run as a standalone python library to extract data lineage from the Fivetran api.

The library is available via pip

pip install grai_source_fivetran

More information about the API is available here.


The library is split into a few distinct functions but if you only wish to extract nodes/edges from Fivetran you can do so as follows:

  from grai_source_fivetran import FivetranIntegration
  from grai_schemas.v1.source import SourceV1
  source = SourceV1(name="my-source", type="my-type")
  fivetran_credentials = {
    "api_key": "my-api-key",
    "api_secret": "my-api-secret"
  integration = FivetranIntegration(source=source, default_namespace="fivetran", **fivetran_credentials)
  nodes, edges = integration.get_nodes_and_edges()

In this case, we are putting all nodes and edges produced by Fivetran in a single namespace. In practice you usually don't want to do this because it will result in overlapping id's. For example, a fivetran connection copying data from a source table my_table to a destination table my_table will result in two nodes with the same id.

To avoid this, you can pass a namespaces parameter to the FivetranIntegration constructor which will map Fivetran connection id's to source and destination grai namespaces.

    namespaces = {<fivetran_connection_id>: {
        'source': [source_namespace],
        'destination': [destination_namespace]}
    integration = FivetranIntegration(source=source, namespaces=namespaces, **fivetran_credentials)

In order to build the namespaces you'll need to know the fivetran_connection_id for each connection. You can find these in the Fivetran dashboard or using Grai

    integration = FivetranIntegration(source=source, default_namespace="fivetran", **fivetran_credentials)

Remember, Fivetran connectors sync data from a source (like Postgres) to a destination (like Snowflake). In order for the lineage graph to be complete, you'll need to specify the namespace you've chosen in Grai for each.

For a single a Fivetran connector with with the id crunchy-muffin syncing data from a MySQL database you had added to Grai under the prod namespace to a Snowflake database you had added to Grai under the warehouse namespace the namespaces would look like this:

    namespaces = {
        "crunchy-muffin": {
            "source": "prod",
            "destination": "warehouse"