Amazon S3 annotations: Attach rich, queryable context directly to objects

Amazon announced annotations, a new metadata feature for Amazon S3 that lets you attach rich business context directly to objects. Users can store up to 1,000 named annotations per object, with each annotation supporting JSON, XML, YAML, plain text formats up to 1MB in size (1GB total per object).

This feature is designed to support AI agents and autonomous workflows that need metadata that evolves with the data and scales accordingly. Annotations automatically move together with objects during copy, replication, and cross-region transfer, and can be queried through analytics engines such as Amazon Athena.

Compared to existing S3 metadata approaches, annotations offer the following distinctions:

• System-defined metadata: manages fixed attributes such as size and storage class
• User-defined metadata: small custom information under 2KB set at upload time
• Object tags: up to 10 tags for access control and lifecycle management
• Annotations: variable, queryable rich context up to 1GB in size

In particular, with the S3 Tables MCP server, AI models can query annotations in natural language to retrieve data. Additionally, you can instantly query annotations on objects in any storage class without restoring the object or paying data retrieval costs.