AWS Strands + Amazon Transform: How to Modernize 100 Repos Without a Migration Team
Cloud Platform EngineeringRuntime upgrades, SDK migrations, and framework refactors are the kind of engineering work that requires doing once per repo, multiplied across every repo in the portfolio. AWS has built a three-service stack to automate the pattern at scale — and the architecture is worth understanding before the next migration sprint.
Three Services, One Workflow
The AWS agentic modernization stack combines three components, each handling a distinct part of the workflow.
Strands Agents SDK is a multi-agent coordination framework. In the modernization use case, it orchestrates the analysis workflow: inspecting repositories to identify language runtimes, dependency versions, and required transformation types. Where an existing transformation exists in a central registry, Strands retrieves it. Where none exists, it invokes an AI-generation step to create one, then publishes it back to the registry for reuse.
Amazon Transform Custom handles execution. Once a transformation is mapped or generated, Transform Custom runs it via AWS Batch — a CLI-driven invocation that applies the code change to each repository in parallel. The Batch job model is what makes this scale: repositories are processed concurrently rather than queued through a single transformation process.
Amazon Bedrock AgentCore provides the managed runtime layer: workflow state, agent memory, orchestration, and observability across the full transformation run. Without AgentCore managing state, a multi-repo modernization job across dozens of parallel Batch executions becomes difficult to monitor and resume after failure.
What Gets Automated and What Doesn’t
The workflow handles everything from repository discovery to parallel code transformation automatically. Developers submit a batch of repository paths (via CSV or manual input), optionally describe required transformations in natural language, and AgentCore manages the rest: analysis, transformation mapping, registry publication, Batch invocation, and status tracking.
What remains manual: the initial repository submission, validation of generated transformations before they enter the registry, and any code review required after the automated changes are applied. The system is not fully autonomous — it accelerates the preparation and execution phases while preserving human review at merge time.
This positions the stack as a force multiplier for platform teams running language runtime upgrades (e.g., Java 11 to Java 21, Python 3.9 to 3.12) or SDK migrations (e.g., AWS SDK v2 to v3) across large portfolios. The AWS Transform service has been applied to enterprise Java and mainframe modernization; the Strands integration extends that capability to teams managing heterogeneous language portfolios.
The Pattern Emerging in AWS Agentic Tooling
Over the past two quarters, AWS has shipped an interconnected set of agentic services targeting platform teams: Strands for agent orchestration with token-aware tool design, BedrockAgentCore for managed agent runtime, AgentCore payments for autonomous API access, and now Transform Custom for codified, repeatable transformation execution. The pattern is a vertically integrated agentic pipeline — analysis, decision, execution, and observability — all within the AWS boundary.
The enterprise appeal is clear: teams already running workloads on AWS can adopt these services incrementally without multi-vendor orchestration complexity.
The So What
Application modernization is a tax that compounds with technical debt: the longer a runtime upgrade is deferred, the more dependencies need simultaneous updates. The Strands + Transform + AgentCore stack addresses the execution bottleneck — not the decision about what to modernize, but the throughput constraint in doing it. Platform engineers who have spent sprints on repo-by-repo Java upgrades will recognize what parallel Batch-driven transformation means for timeline. The stack is available now; the question is whether the transformation registry in your organization is mature enough to feed it.
Photo by Digital Buggu on Pexels
Source: AWS DevOps Blog
Content created with AI assistance and reviewed for accuracy.
Join the conversation
Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.
Join Stack Insiders →