aws-agentcore-langgraph
por killerapp
DevOps & CloudResumen
Deploy production LangGraph agents on AWS.
Descripción
Qué hace esta skill
Deploys LangGraph agents to AWS Bedrock AgentCore with support for multi-agent orchestration, persistent memory across sessions, and external tool integration via Gateway. Wraps LangGraph state graphs as HTTP services running on port 8080. Includes a CLI for configuration, deployment, and lifecycle management.
Combines LangGraph's stateful graph routing with AWS-managed memory, scaling, and tool connectivity in one deployment target, avoiding the need to wire these infrastructure pieces together manually.
Cuándo usarla
- Deploying a customer service agent ecosystem with specialist sub-agents
- Adding cross-session memory to a LangGraph chatbot so it recalls past decisions
- Exposing Lambda functions and REST APIs as MCP tools for agent use
- Running a local hot-reload dev server while iterating on agent logic
- Scripting unattended production deployments via agentcore CLI flags
Flujo de trabajo de ejemplo
User asks: deploy my LangGraph agent with persistent memory to AWS
- 1
Installs bedrock-agentcore and langgraph packages
- 2
Wraps existing StateGraph with BedrockAgentCoreApp and defines an entrypoint
- 3
Runs agentcore configure -e agent.py --region us-east-1 to set up deployment config
- 4
Runs agentcore launch --deployment-type container to build and deploy
- 5
Runs agentcore invoke '{"prompt": "Hello"}' to verify the deployed endpoint responds
Agent running as a managed AWS service with /invocations and /ping endpoints, memory subsystem enabled, ready to handle production traffic
Requisitos
- AWS account with Bedrock access enabled
- AWS credentials configured (via environment variables or ~/.aws/credentials)
- Anthropic model access approved in AWS Bedrock Console (requires submitting model use case form)
Instalación
clawhub install killerapp/aws-agentcore-langgraphEjecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.