Amazon SageMaker Unified Studio Workflows now supports operators for Amazon Bedrock, S3 Tables, S3 Vectors, and Glue Catalog
19 new no-code operators let you orchestrate Bedrock, S3 Vectors, Glue, and MWAA together — without writing a single line of DAG code.
View original announcement →Visual Summary
graph TD
A{{SageMaker Unified Studio Workflows}}:::announced
B((Data Engineers)):::external
C([Visual Workflow Creator]):::feature
D(Amazon Bedrock):::compute
E(Amazon S3 Tables):::storage
F(Amazon S3 Vectors):::storage
G(AWS Glue Data Catalog):::storage
H(MWAA Serverless):::compute
I([19 New Operators]):::feature
B ==>|"drag-and-drop"| A
A ==>|"provides"| C
C -->|"generates"| I
I -->|"manages guardrails"| D
I -->|"provisions"| E
I -->|"provisions"| F
I -->|"manages tables"| G
I -.->|"triggers DAG runs"| H
classDef announced fill:#ff9900,stroke:#ec7211,color:#fff,font-weight:bold
classDef compute fill:#e3f2fd,stroke:#1565c0,color:#1565c0
classDef storage fill:#e8f5e9,stroke:#2e7d32,color:#2e7d32
classDef feature fill:#fff3e0,stroke:#e65100,color:#e65100
classDef external fill:#f5f5f5,stroke:#616161,color:#616161
What's New
Amazon SageMaker Unified Studio Workflows has added 19 new Apache Airflow-compatible operators spanning Amazon Bedrock, Amazon S3 Tables, Amazon S3 Vectors, AWS Glue Data Catalog, and Amazon MWAA Serverless. These operators are accessible through the visual workflow creator's drag-and-drop interface, enabling users to orchestrate these services without writing custom DAG code or integration logic. The expansion brings the total supported task count to over 80, covering analytics, compute, catalog, and storage use cases across the AWS ecosystem.
How It Works
- The visual workflow creator in SageMaker Unified Studio provides a drag-and-drop canvas where users select operators from a task library and connect them into a directed acyclic graph (DAG) representing execution order and data flow.
- New Amazon Bedrock operators allow workflows to manage Bedrock guardrails directly, enabling governance and safety configurations for generative AI pipelines to be embedded as workflow steps.
- New Amazon S3 Tables and S3 Vectors operators support provisioning and deletion of these storage resources, allowing data lifecycle management to be automated within a single workflow definition.
- New AWS Glue Data Catalog operators enable creation, modification, and deletion of Glue tables and databases, so schema and metadata management can be orchestrated alongside data processing tasks.
- New Amazon MWAA Serverless operators allow workflows to trigger external MWAA Serverless DAG runs, enabling cross-workflow orchestration and reuse of existing Airflow pipelines.
- Under the hood, the visual workflow editor generates Apache Airflow DAG code, which can be viewed and exported, giving advanced users full transparency and the ability to extend or migrate workflows.
- Workflows can be scheduled using cron expressions or preset schedules, and support parameterization and Airflow tags for filtering and reuse.
Why It's Important
- Reduced context switching: Data engineers and ML practitioners no longer need to navigate multiple AWS consoles to orchestrate Bedrock, S3 Tables, S3 Vectors, and Glue together — everything is managed from a single unified interface.
- Lower barrier to entry: The no-code visual interface means data workers without deep Airflow expertise can build production-grade orchestration pipelines that previously required custom DAG authoring.
- AI/ML pipeline integration: Native Bedrock guardrail management within workflows enables teams to enforce responsible AI policies as a first-class step in their data and model pipelines, not as an afterthought.
- Modern storage orchestration: With S3 Tables (Apache Iceberg-based) and S3 Vectors (purpose-built for vector embeddings) operators, teams can automate the full lifecycle of next-generation storage formats used in analytics and RAG-based AI applications.
- Unified governance: Glue Data Catalog operators allow metadata and schema governance to be co-located with the data processing logic, improving auditability and reproducibility of data workflows.
- Accelerated time-to-value: Pre-built, tested operators eliminate the need to write and maintain custom Airflow provider code, reducing development time and operational risk.
How It's Different
- Versus raw Apache Airflow on MWAA: Standard MWAA requires users to write Python DAG files and manage provider package dependencies; SageMaker Unified Studio Workflows abstracts this into a visual, code-optional experience while still generating valid Airflow DAGs.
- Versus AWS Step Functions: Step Functions offers visual workflow authoring but is not Airflow-native and lacks the rich ecosystem of Airflow operators; SageMaker Unified Studio Workflows bridges the gap by providing Airflow semantics with a visual interface.
- Versus custom Lambda/Glue orchestration: Hand-rolled orchestration using Lambda or Glue workflows requires significant boilerplate and error-handling code; the new operators provide pre-built, retry-aware task definitions out of the box.
- Versus standalone Bedrock console: Bedrock guardrail management through the console is a manual, one-off action; embedding it as a workflow operator makes it repeatable, auditable, and schedulable.
- Breadth of coverage: With 80+ supported tasks across compute, analytics, storage, catalog, and now generative AI services, SageMaker Unified Studio Workflows offers broader AWS service coverage in a visual tool than competing workflow builders.
- DAG portability: Because the visual editor generates exportable Airflow DAG code, teams are not locked into the visual interface and can migrate or extend workflows in standard Airflow environments.
When to Prefer It
- Building RAG or generative AI data pipelines: When you need to provision S3 Vectors stores, manage Bedrock guardrails, and catalog metadata in Glue as part of a single automated pipeline for retrieval-augmented generation workloads.
- Automating data lakehouse lifecycle management: When your team needs to programmatically create, update, or delete S3 Tables (Iceberg) and Glue Data Catalog entries as part of a scheduled ETL or data ingestion workflow.
- Teams with mixed technical skill levels: When your organization includes both data engineers comfortable with Airflow and business analysts or data scientists who prefer a no-code interface — the visual editor serves both audiences from the same tool.
- Migrating existing Airflow DAGs: When you have existing MWAA DAGs and want to integrate them into a broader SageMaker Unified Studio workflow without rewriting them, using the MWAA Serverless trigger operator.
- Enforcing AI governance at pipeline level: When your organization requires Bedrock guardrail configurations to be applied consistently and automatically as part of model deployment or inference pipeline workflows.
- Reducing multi-console operational overhead: When your team spends significant time switching between SageMaker, S3, Glue, and Bedrock consoles to manage related tasks that logically belong to the same workflow.
- Rapid prototyping of complex orchestration: When you want to quickly assemble and test multi-service workflows without investing time in custom integration code, then optionally export the generated DAG for production hardening.
Availability
- GA Status: Generally available as of July 8, 2026; this is a production-ready feature, not a preview.
- Supported Regions: Available in all AWS Regions where Amazon SageMaker Unified Studio is currently supported; refer to the AWS Regional Services List for the definitive region table.
- Operator Count: 19 new operators added in this release, bringing the total visual workflow task library to over 80 supported tasks.
- Pricing: No separate charge is announced for the operators themselves; costs are incurred by the underlying AWS services invoked (Bedrock, S3, Glue, MWAA Serverless) per their standard pricing.
- Dependency: Requires an active Amazon SageMaker Unified Studio environment and appropriate IAM permissions for each target service (Bedrock, S3, Glue, MWAA).
- Limitations: The feature relies on Amazon MWAA Serverless as the underlying execution engine; operators are limited to the set of services explicitly supported by the platform and cannot yet cover all AWS services.
- Documentation: Full operator reference available at the MWAA Serverless User Guide operators page; getting-started guide available in the SageMaker Unified Studio User Guide under "Serverless visual workflows."
