Developers can now deploy a local MCP server to AWS in just two commands: `agentcore configure` and `agentcore launch`. Using the Bedrock AgentCore Starter Toolkit, the entire workflow—from creating a simple dice-rolling server with FastMCP to testing locally—takes only minutes. The toolkit automatically generates the necessary Dockerfiles, IAM roles, and infrastructure code for deployment.

AgentCore Runtime is specifically designed for agent workloads, solving the serverless state problem. Unlike AWS Lambda, it maintains session state across multiple calls while keeping each session in a completely isolated environment. This eliminates the need to manage state via DynamoDB or run costly dedicated containers, offering a pay-for-processing model that's ideal for agentic applications.

The deployment process uses CodeBuild for cloud-based container builds, meaning developers don't need Docker locally. After deployment, servers are accessible via IAM credentials, bypassing complex OAuth setups. The AWS CLI's `invoke-agent-runtime` command allows testing, and the IAM MCP Client integrates with frameworks like Strands Agents for real-world use. This streamlined path from local development to production is a significant step for the MCP ecosystem.