Global – October 28, 2026 -- Snowflake has made automated Snowpipe setup with its CoCo assistant generally available, letting teams take raw files in cloud storage to a continuously loading Snowflake table or Snowflake-managed Apache Iceberg table in a single natural-language conversation.
CoCo samples the files, infers the schema, builds the pipeline and confirms that data loads. The feature is available in Snowsight, CoCo Desktop and CoCo CLI, and anyone with the right permissions can connect a new source without waiting in a data engineering queue.
Snowflake removes the manual steps that delay the first load
Snowpipe, Snowflake's continuous loading service, typically picks up each new file within minutes of arrival on Snowflake-managed serverless compute. The delay comes earlier. Someone must inspect files to choose format options, then create the storage integration, stage, table and pipe. The cloud team must also configure an IAM role, a trust policy and event notifications.
CoCo points at Amazon S3, Azure Blob Storage or Google Cloud Storage paths. It detects formats including CSV, JSON, Parquet, Avro, ORC and XML, and proposes column names and data types. It reuses compatible storage integrations, stages and file formats, and flags any existing pipe that already covers the same files to avoid double loading.
AWS IAM temporary delegation limits access to one approved, time-bound request
On Amazon S3, users approve a single time-limited AWS IAM temporary delegation request in the AWS Management Console, with no AWS access keys to share. The request covers notification access on one bucket and permission to create or update the storage role, capped by a permissions boundary. Snowflake then creates the storage role and trust policy and adds the pipe's notification rule, leaving the bucket's existing rules in place.
AWS CloudTrail logs every setup action taken under the delegated role. Snowflake query history records each statement CoCo runs under the user's role.
"Every project still starts with connecting new data," said Saptarshi Mukherjee in the announcement.
Four setup paths cover Terraform, existing access and generated cloud files
CoCo offers four options: temporary access on Amazon S3, Terraform for the complete pipeline, existing cloud access, or Snowflake objects plus cloud setup files. For standard and Iceberg tables on AWS, Azure and Google Cloud, CoCo writes Terraform covering both Snowflake objects and cloud resources, ready for pull-request review. The setup-file option generates Terraform, CloudFormation or ARM templates for cloud teams.
CoCo Desktop and CoCo CLI can apply the files using local cloud credentials, asking before each application. By default, CoCo shows the SQL and requests approval before creating or changing any Snowflake object.
Backfill covers seven days, and Iceberg output opens data to Spark
CoCo offers to load the last seven days of files that landed before the pipe existed, and can run a one-time load for older data. It can enable schema evolution so new source fields become new columns, asking first on an existing table.
For Iceberg targets, Snowpipe writes data as Parquet, converting other formats on load. Snowflake maintains the Iceberg metadata and table maintenance, so Apache Spark and other Iceberg-compatible engines can read the same table.
After setup, CoCo checks pipe status, notification delivery and load history and reports load errors, replacing manual queries of SYSTEM$PIPE_STATUS or COPY_HISTORY. Snowflake also made bulk loading with CoCo generally available. It prepares the file format and table, runs COPY INTO on the user's warehouse and checks results for backfills and migrations.
Snowpipe loads run on serverless compute and are billed as Snowpipe usage.