Amazon S3 annotations: attach rich, queryable context directly to your objects

Amazon S3 just added a feature called annotations that fundamentally changes how you can work with object metadata at scale. Instead of managing metadata in separate databases or systems, you can now attach up to 1 GB of rich, queryable context directly to S3 objects. For teams building AI agents and automation workflows, this is a practical shift that simplifies data discovery and context management in ways that single-key/value tag systems simply can’t match.

Here’s how it works technically. Annotations are flexible JSON documents attached to objects that live alongside your data in S3. Unlike traditional S3 tags (which have significant limitations on size and queryability), annotations support structured data, nested objects, and arrays—essentially anything valid JSON can represent. You query them through the S3 Select API and other S3 operations, meaning your applications can discover and filter objects based on rich context without pulling data into separate metadata stores. This matters because modern AI workflows often need to understand not just what an object is, but semantic information about it: who processed it, what quality checks it passed, what transformations it’s undergone, or what downstream systems depend on it.

Consider a practical example: an insurance company processing claim documents. When a document enters S3, an AI agent can attach annotations containing extracted claim details, quality scores, compliance flags, and routing information—all queryable without leaving S3. Another workflow searching for “high-confidence claims from Q4 that passed fraud checks” can filter objects by those annotations directly. Previously, you’d need DynamoDB or a database running parallel to your S3 bucket, syncing metadata constantly and introducing potential inconsistencies. With annotations, the metadata stays with the object, versioning automatically, and available for any service that needs it.

For growing teams, the practical benefit is reducing operational complexity. You’re eliminating the need to maintain separate metadata infrastructure, reconcile sync issues between systems, and design schemas specifically for “what metadata do we need to track?” Autonomous workflows can be more self-contained, and new use cases don’t require database migrations. It’s designed for the kind of scale where you might have millions of objects flowing through AI pipelines daily, and each one needs to carry rich context without creating bottlenecks. If you’re already managing complex workflows in S3, it’s worth exploring whether annotations could simplify your metadata story.