Snowflake makes CoCo’s automated Snowpipe setup generally available

Snowflake has made automated Snowpipe setup with CoCo generally available, allowing users to build continuous ingestion pipelines from cloud storage with a plain-language request. The October 27 release is available in CoCo CLI, CoCo Desktop and CoCo in Snowsight.

The workflow supports Amazon S3, Microsoft Azure Blob Storage and Google Cloud Storage. It can target standard Snowflake tables or Snowflake-managed Apache Iceberg tables. CoCo samples source files, infers the schema or asks users to confirm it, selects file-format settings, and then creates resources or generates deployment files for the chosen setup path.

Snowpipe uses cloud event notifications to detect new files and runs the COPY INTO statement stored in a pipe object on Snowflake-managed serverless compute. Snowflake says CoCo replaces the manual coordination of storage and notification integrations, stages, file formats, destination tables and pipes with a guided setup conversation. The underlying Snowpipe pipeline continues loading after the CoCo session ends.

Snowflake documents four setup paths. Teams can give Snowflake temporary access to automate AWS cloud setup, generate Terraform for the complete cloud and Snowflake pipeline, reuse existing cloud access, or have CoCo create Snowflake objects and generate cloud deployment files. The Terraform path is available for AWS, Azure and Google Cloud, including Snowflake-managed Iceberg pipelines.

The required approvals vary by path and client. The temporary-access option is limited to Amazon S3 and is unavailable in AWS GovCloud and China deployments. Before Snowflake can create the IAM role, connect bucket notifications and create Snowflake objects, a user or AWS administrator must approve a temporary delegation request in the AWS Management Console. For Terraform or other generated deployment files, CoCo CLI and Desktop apply the files only with user approval; otherwise, the user or their team deploys them. CoCo in Snowsight saves the files to a workspace but does not apply them.

CoCo displays the full SQL before changing Snowflake objects. By default, it waits for approval before each DDL statement, though users can bypass those SQL approvals. Cloud authorization and deployment approvals still apply to the chosen setup flow. CoCo also checks for overlapping pipes before proceeding with a configuration that could load the same files twice. After access and deployment are complete, it checks pipe status, notification delivery and load history, verifies that arriving files load, and surfaces errors.

For Snowflake-managed Iceberg targets, CoCo selects the load mode from the input. It uses ADD_FILES_COPY only for Parquet files produced by an Iceberg writer. Other Parquet files, along with CSV, JSON, Avro and ORC files, use FULL_INGEST. Complete-pipeline Terraform creates the table with Snowflake as its catalog. AWS and Azure accounts default to Snowflake storage, while Google Cloud, government-region and China deployments require an external volume. This Terraform path does not support Iceberg tables that use an external catalog.

New-file notifications do not automatically ingest files already present at the source. CoCo asks whether to load files staged within the previous seven days; older files require a separately planned historical load.