AWS details four-agent pattern for large cloud migrations on Bedrock AgentCore

AWS has described a four-agent pattern on Amazon Bedrock AgentCore for large cloud migrations. According to AWS, it reduced infrastructure as code (IaC) development time from 3 to 4 weeks per application to minutes on an enterprise program covering 300+ applications, based on internal project tracking data.
- AWS describes a four-agent pattern on Amazon Bedrock AgentCore for large cloud migrations.
- According to AWS, IaC development time fell from 3 to 4 weeks per application to minutes on a 300+ application program.
- The agents run alongside AWS Transform and AWS DMS rather than replacing them.
- The agents are Intake, IaC, Migration Intelligence and Governance, and SRE, connected via MCP tools through AgentCore Gateway.
- The post includes Python code showing an agent with Amazon Bedrock Guardrails attached.
A hybrid with AWS managed services
The pattern is not meant to replace AWS Transform. AWS says it runs alongside the service as a hybrid that adds custom agents where a program requires them. AWS Transform covers migration and modernization work, and AWS Database Migration Service (AWS DMS) covers the database tier.
The agents attach to those services and address one further requirement: sources and destinations reached through Model Context Protocol (MCP) tools that an organization builds and maintains. AWS Professional Services builds the suite of purpose-built agents for programs with that requirement. They use the Strands Agents SDK and run on Amazon Bedrock AgentCore, which AWS describes as a platform to build, connect, and optimize agents at scale, with any framework or model.
AWS advises readers to confirm first that a managed service does not already cover their migration path. It says AWS Transform handles server, network, mainframe, .NET, and application code workloads, while AWS DMS handles databases.
The four agents
The pattern splits the work into a migration journey and an operations journey.
- Intake Agent (Phase 1): reads architecture documents, questionnaires, and dependency records through MCP tools, then defines the target state architecture.
- IaC Agent (Phase 2): generates IaC that composes the organization's approved internal modules for each application.
- Migration Intelligence and Governance Agent: provides automated portfolio reporting, well-architected assessments, and governance across Jira, Confluence, and Webex.
- SRE Agent (Phase 3): provides monitoring and automated remediation after cutover.
AWS says three conditions held together on the program it describes. The systems holding inputs and receiving outputs were reached through MCP tools built by the delivery team, including an internal wiki with security standards, a ticketing system, a collaboration platform, and an in-house provisioning API. Generated code had to compose an internal module library reviewed and approved by the security office. And the scope included operations after handover, which sits outside the migration services. According to AWS, writing the IaC composition by hand took 3 to 4 weeks per application, which across 300+ applications translates to years of engineering effort.
How the components connect
Each agent is a Strands agent defined by a foundation model, a system prompt, and a set of tools. Amazon Bedrock AgentCore runtime hosts them in a serverless environment with session isolation and multi-agent orchestration, while Amazon Bedrock foundation models provide the reasoning for interpreting documents, generating code, and driving multi-step workflows.
Each agent calls MCP tools scoped to its function through AgentCore Gateway, which converts APIs, AWS Lambda functions, and existing services into MCP-compatible tools. AgentCore Identity authenticates each call through scoped AWS Identity and Access Management (IAM) roles and the organization's identity provider. AgentCore memory stores session state and shared context, so agents can track progress across the 300+ applications.
That shared memory removes manual handoffs. When the Intake Agent finishes discovery, it writes the target architecture and dependency mappings to memory, and the IaC Agent reads them to begin code generation.
Code and responsible AI controls
The post includes a Python example that defines the IaC Agent for the AgentCore runtime. The agent reaches MCP tools through AgentCore Gateway and calls a foundation model through Amazon Bedrock with an Amazon Bedrock Guardrails policy attached. According to the code comments, it fetches the approved policies for a migration wave first, so the rules travel in the system prompt rather than depending on the model to request them. If a guardrail intervenes, the request is reported as blocked.
To follow along, readers need an AWS account with access to Amazon Bedrock AgentCore and Amazon Bedrock foundation models, plus familiarity with the Strands Agents SDK, MCP server patterns, and their organization's IaC tooling.
What it means for migration teams
The post suggests a division of labor: use managed services such as AWS Transform and AWS DMS where they cover the path, and add custom agents only for requirements specific to the organization, such as internal module libraries and MCP-connected internal systems. The time savings cited come from AWS's own project tracking data on a single program, so results elsewhere may differ.


