Agent Builder vs Bedrock Agent use case
This solution provides two distinct approaches for working with AI agents, each suited for different use cases and requirements:
| Feature | Bedrock Agent use case | Agent Builder |
|---|---|---|
|
Purpose |
Invoke pre-deployed Amazon Bedrock Agents |
Build, deploy, and manage custom agents |
|
Configuration |
Agent ID and Alias ID only |
Full agent configuration: system prompts, models, MCP servers, memory |
|
Deployment |
Simple invocation layer |
Complete agent lifecycle on AgentCore Runtime |
|
Runtime |
Amazon Bedrock Agents service |
Amazon Bedrock AgentCore with Strands SDK |
|
Tool Integration |
Configured in Bedrock Agents console |
Model Context Protocol (MCP) servers and built in Strands tools |
|
Memory |
Managed by Bedrock Agents (up to 30 days) |
AgentCore Memory with configurable short-term and long-term retention |
|
Customization |
Limited to pre-deployed agent settings |
Full control over prompts, models, tools, and behavior |
|
Best for |
Quick deployment of existing agents |
Custom agent development and production deployments |
Note
Both options support real-time streaming, conversation history, and enterprise-grade security.
