The platform layer provides the foundational infrastructure for building, deploying, and operating multi-agent AI systems on AWS.
The platform is designed to be:
- Framework-agnostic — supports LangGraph, LangChain, Strands, and custom frameworks
- Deployment-ready — deploy to Amazon Bedrock AgentCore with built-in auto-scaling and observability
- Scalable — shared infrastructure with per-use-case isolation via Terraform workspaces
The AgentCore adapter translates between Bedrock AgentCore Runtime and the agent registry:
| Adapter | File | Purpose |
|---|---|---|
| AgentCore | foundations/src/adapters/agentcore_adapter.py |
Bedrock AgentCore Runtime deployment |
Central registration point for all use cases. The registry (foundations/src/base/registry.py) allows any adapter to invoke any registered agent:
from base.registry import register_agent, RegisteredAgent
register_agent("kyc_banking", RegisteredAgent(
entry_point=run_kyc_assessment,
request_model=KYCRequest,
response_model=KYCResponse,
))Framework-specific base classes for building agents and orchestrators:
foundations/src/base/langgraph/— State-based workflow orchestrationfoundations/src/base/langchain/— Chain-based agent compositionfoundations/src/base/strands/— AWS Strands SDK integration
| Pattern | Infrastructure | Best For |
|---|---|---|
| AgentCore | AWS-native serverless | Scalable, managed deployments |
See Architecture Patterns for detailed information.
| Variable | Description | Default |
|---|---|---|
AWS_REGION |
AWS region | us-east-1 |
S3_BUCKET_NAME |
Data bucket | Required |
BEDROCK_MODEL_ID |
Bedrock model | Claude Sonnet |
AGENT_NAME |
Agent to run | kyc_banking |
DEPLOYMENT_PATTERN |
Deployment pattern | agentcore |
pip install -r foundations/src/requirements/requirements.txt
export AWS_REGION=us-east-1
export S3_BUCKET_NAME=your-bucket
export AGENT_NAME=kyc_banking
export DEPLOYMENT_MODE=local
cd foundations/src
python main.py# Interactive deployment
./scripts/main/deploy.sh
# Or deploy a specific pattern
./scripts/deploy/full/deploy_agentcore.sh- AgentCore Architecture — Bedrock AgentCore details
- Architecture Patterns — Deployment architecture
- AgentCore Deployment — AgentCore deployment guide
- Deployment Patterns — Step-by-step deployment
- Security Architecture — Security architecture and data flows
- Threat Model — Security threat analysis and mitigations
- AWS Service Security — Service-specific security guidelines
- AI Security — GenAI-specific security considerations
- Adding Applications — Create new use cases
- Local Development — Development setup
- Global Variables — Environment variable reference