Exam Room · Advanced Generative AI Developer

Pop Quiz: Strands, Agent Squad, or AgentCore

· 5 min read

Exam-style

A support team already runs a single Strands agent on the Amazon Bedrock AgentCore runtime. It has now split that agent into four specialists, for billing, delivery, returns and account changes, and needs a front door that sends each incoming message to the right specialist and keeps the thread coherent when a subscriber changes topic halfway through. What should the team add?

Reveal the answer

C. Add AWS Agent Squad in front as the classifier-based router, leaving the four specialists built on Strands Agents and hosted on the AgentCore runtime

The three names sit at three different layers, and the guide keeps naming them in one breath, which makes them look like alternatives. Strands Agents is the framework you build an agent in: the model, the system prompt, the tool definitions, the loop. AWS Agent Squad is the layer above, the AWS native orchestration piece for multi-agent systems, It classifies each incoming message against the descriptions of the agents registered with it and routes to the one that fits. It also carries shared conversation state, so a topic change mid-thread lands on a different specialist without losing what came before. AgentCore is the runtime underneath, which hosts whatever you built and gives it session isolation, identity and memory. Adding a router changes none of that: the specialists stay as Strands agents, they stay on the AgentCore runtime, and MCP for agent-tool interactions still carries every tool call. Collapsing the four back into one agent throws away the separation that made routing tractable. It also puts every tool in one context window. Rebuilding the specialists on Agent Squad misreads it as a framework. A Step Functions Choice state can branch a workflow. A keyword match is not intent classification, and it has nothing useful to say about a subscriber who opens with a delivery complaint and ends up asking to close the account. AgentCore Memory stores conversation state for an agent; it does not choose which agent runs.

Generative AI Development · part of The Exam Room

Q. Four specialist agents and a front door that has to pick one per message. Strands, Agent Squad, or AgentCore?

A. All three, at their own layers. Add AWS Agent Squad as the classifier-based router in front; keep the four specialists built on Strands Agents and running on the AgentCore runtime.

Why? Strands builds an agent, Agent Squad routes between agents, AgentCore runs them, and MCP is how any of them reaches a tool. Reaching for one does not replace the others, so a routing problem gets solved by adding the routing layer rather than rewriting what already works.

These posts are LLM-aided. Backbone, original writing, and structure by Craig. Research and editing by Craig + LLM. Proof-reading by Craig.