AWS Agentcore

AWS Bedrock AgentCore

Overview

Build and deploy AI agents using AWS Bedrock AgentCore. Supports Strands, Claude, OpenAI SDKs and Bedrock/OpenAI model providers. Infrastructure via CDK or Terraform.

MCP Tools

• search_agentcore_docs - Search AgentCore documentation
• fetch_agentcore_doc - Retrieve specific doc pages
• manage_agentcore_runtime - Runtime configuration and deployment
• manage_agentcore_memory - Agent memory operations
• manage_agentcore_gateway - Gateway configuration

Getting Started

1. Use agentcore create for new agent projects
2. Run agentcore dev for local development with hot reloading
3. Test locally before cloud deployment
4. Deploy with agentcore deploy

Troubleshooting

• "Could not find entrypoint module" → Ensure .bedrock_agentcore.yaml exists
• "Port 8080 already in use" → CLI auto-tries next port
• "AWS authentication failed" → Run aws login
• "Model access denied" → Check Bedrock model permissions

MCP Server Config

{
  "mcpServers": {
    "agentcore": {
      "command": "uvx",
      "args": ["awslabs.agentcore-mcp-server@latest"]
    }
  }
}