Agentic Frameworks on AWS
A collection of AWS reference examples and blog posts showing how to build autonomous agent applications on AWS using open-source frameworks such as LangGraph, CrewAI and LlamaIndex, spanning foundation models, orchestration, memory, tools and evaluation.
This repository provides example architectures for building production-oriented agent applications on AWS services using popular open-source agentic frameworks, aimed at demonstrating patterns across the layers of an agent stack: foundation models, orchestration, memory, tools, observability and evaluation, and deployment. The example projects cover an agent-memory pattern for an insurance domain, an end-to-end serverless multi-agent system built on the Agent2Agent protocol, an AWS infrastructure security-audit and reporting agent crew, automating customer support with Amazon Bedrock and LangGraph using Mistral models, multi-agent collaboration and orchestration with LangGraph, a LangChain-based router agent across the Mistral and AWS ecosystem, and a vision question-answering agent built with Mistral and LlamaIndex.
Linked AWS blog posts extend the same themes to CrewAI, DeepSeek-R1 on SageMaker, extending Bedrock-hosted models with the Model Context Protocol, and automating regulatory compliance with a multi-agent system, and linked workshops cover observability and evaluation for agents with Langfuse, LangGraph agents on Bedrock, and building effective agents on AWS more generally.