Amazon Web Services has expanded its Amazon Bedrock AgentCore Runtime with new stateful MCP client capabilities. These allow developers to build servers that maintain context during interactions. The update enables servers to request user input mid-execution. It also supports dynamic content generation through LLM sampling. Progress updates can now stream in real time for long-running tasks.
A new guide provides code examples for each feature. It covers stateful server setup, user input handling, and progress streaming. The documentation includes a step-by-step deployment example. Developers can now run a working stateful MCP server on Amazon Bedrock AgentCore Runtime.
The feature targets applications needing persistent context. Examples include multi-turn conversations and workflow automation. AWS states this reduces the need for external state management. The company highlights improved developer experience for complex agentic systems.
Documentation and code samples are available in the AWS Machine Learning Blog. The update follows AWS’s push to expand agentic AI tools. It aligns with broader industry trends toward stateful AI interactions.
Source: aws.amazon.com