Snowflake Automates Snowpipe Setup with CoCo

Snowflake announced general availability of automated Snowpipe setup with Snowflake CoCo on October 27, 2026. Teams can describe a cloud storage source in a natural-language conversation, and CoCo samples files, infers a schema, builds the pipeline and checks that data loads into a standard Snowflake table or Snowflake-managed Apache Iceberg table. The feature is available in Snowsight, CoCo Desktop and CoCo CLI. By default, CoCo asks users to approve changes before making them; once setup is complete, Snowpipe continues loading after the CoCo session ends.

Snowpipe is Snowflake’s continuous loading service, which typically picks up each new file within minutes of its arrival using serverless compute managed by Snowflake. Before this launch, teams needed to inspect files and create a storage integration, stage, table and pipe. For AWS sources, the cloud team also had to configure an IAM role, trust policy and file arrival notifications. CoCo can connect sources from Amazon S3, Azure Blob Storage and Google Cloud Storage, detect formats including CSV, JSON, Parquet, Avro, ORC and XML, and propose column names and data types. It can reuse compatible integrations, stages and file formats; if a pipe already covers the files, it flags the overlap to avoid loading them twice.

For Amazon S3 setup, CoCo can request one-time, time-limited AWS IAM delegation. Users review the requested permissions in the AWS Management Console and can approve, deny or send the request to an administrator. Snowflake then creates the approved storage role and trust policy; after the user approves the SQL that creates the pipe, Snowflake adds its notification rule while preserving existing bucket rules. AWS CloudTrail records setup actions under the delegated role, and Snowflake query history records CoCo statements under the user’s role. The temporary access expires; the storage role and bucket notifications remain for ongoing loads. AWS IAM temporary delegation is unavailable in AWS GovCloud (US) and China regions.

Teams can instead have CoCo generate Terraform for the complete pipeline, including Snowflake objects and cloud resources, on AWS, Azure or Google Cloud. The team can review the code in a pull request and deploy it through its usual process. Other choices include reusing existing cloud access, or having CoCo create approved Snowflake objects and produce Terraform, CloudFormation or ARM templates for the cloud team. CoCo Desktop and CoCo CLI can apply these files using credentials on the user's machine, with approval before each application; Snowsight saves them in a workspace folder. Terraform generated for AWS preserves existing bucket notification rules.

For an Iceberg destination, CoCo creates a Snowflake-managed Iceberg table and its loading pipe. Snowpipe writes data as Parquet, converting other file formats during loading, while Snowflake maintains the Iceberg metadata and table maintenance. Apache Spark and other Iceberg-compatible engines can read the table. CoCo checks pipe status, notification delivery and load history, reports errors and can guide users through a test file if no new files have arrived. It also offers to load files from the previous seven days immediately and can run a one-time load for older files; schema evolution can add new fields as columns, with approval requested for an existing table.

Bulk loading with CoCo is also generally available. For files already in place, users can request a one-time load; CoCo prepares the file format and table, runs COPY INTO on a warehouse and checks the results. If the request does not specify one-time or continuous loading, CoCo asks which is intended. Snowpipe usage is billed for its serverless compute. Cloud approvals and permission propagation can affect setup time.

Snowflake announced general availability of automated Snowpipe setup with Snowflake CoCo on October 27, 2026. Teams can describe a cloud storage source in a natural-language conversation, and CoCo samples files, infers a schema, builds the pipeline and checks that data loads into a standard Snowflake table or Snowflake-managed Apache Iceberg table. The feature is available in Snowsight, CoCo Desktop and CoCo CLI. By default, CoCo asks users to approve changes before making them; once setup is complete, Snowpipe continues loading after the CoCo session ends.

Snowpipe is Snowflake’s continuous loading service, which typically picks up each new file within minutes of its arrival using serverless compute managed by Snowflake. Before this launch, teams needed to inspect files and create a storage integration, stage, table and pipe. For AWS sources, the cloud team also had to configure an IAM role, trust policy and file arrival notifications. CoCo can connect sources from Amazon S3, Azure Blob Storage and Google Cloud Storage, detect formats including CSV, JSON, Parquet, Avro, ORC and XML, and propose column names and data types. It can reuse compatible integrations, stages and file formats; if a pipe already covers the files, it flags the overlap to avoid loading them twice.

For Amazon S3 setup, CoCo can request one-time, time-limited AWS IAM delegation. Users review the requested permissions in the AWS Management Console and can approve, deny or send the request to an administrator. Snowflake then creates the approved storage role and trust policy; after the user approves the SQL that creates the pipe, Snowflake adds its notification rule while preserving existing bucket rules. AWS CloudTrail records setup actions under the delegated role, and Snowflake query history records CoCo statements under the user’s role. The temporary access expires; the storage role and bucket notifications remain for ongoing loads. AWS IAM temporary delegation is unavailable in AWS GovCloud (US) and China regions.

Teams can instead have CoCo generate Terraform for the complete pipeline, including Snowflake objects and cloud resources, on AWS, Azure or Google Cloud. The team can review the code in a pull request and deploy it through its usual process. Other choices include reusing existing cloud access, or having CoCo create approved Snowflake objects and produce Terraform, CloudFormation or ARM templates for the cloud team. CoCo Desktop and CoCo CLI can apply these files using credentials on the user's machine, with approval before each application; Snowsight saves them in a workspace folder. Terraform generated for AWS preserves existing bucket notification rules.

For an Iceberg destination, CoCo creates a Snowflake-managed Iceberg table and its loading pipe. Snowpipe writes data as Parquet, converting other file formats during loading, while Snowflake maintains the Iceberg metadata and table maintenance. Apache Spark and other Iceberg-compatible engines can read the table. CoCo checks pipe status, notification delivery and load history, reports errors and can guide users through a test file if no new files have arrived. It also offers to load files from the previous seven days immediately and can run a one-time load for older files; schema evolution can add new fields as columns, with approval requested for an existing table.

Bulk loading with CoCo is also generally available. For files already in place, users can request a one-time load; CoCo prepares the file format and table, runs COPY INTO on a warehouse and checks the results. If the request does not specify one-time or continuous loading, CoCo asks which is intended. Snowpipe usage is billed for its serverless compute. Cloud approvals and permission propagation can affect setup time.