itinai content

←back to Blog

AWS Open-Sources an MCP Server for Bedrock AgentCore to Streamline AI Agent Development

«`html

Understanding the Target Audience

The target audience for the AWS Open-Sourced Model Context Protocol (MCP) Server includes software developers and data scientists focused on AI agent development. These professionals aim to streamline their development processes and improve the efficiency of their workflows.

  • Pain Points:
    • Complexity in deploying AI agents due to cloud-specific knowledge requirements.
    • Time-consuming integration of multi-step processes and environment provisioning.
    • Lack of accessible resources for transitioning existing code to new frameworks.
  • Goals:
    • To simplify the process of developing and deploying AI agents.
    • To enhance productivity through tools that reduce manual coding and streamline workflows.
    • To leverage existing knowledge while adapting to new technologies.
  • Interests:
    • New open-source tools and frameworks that enhance AI development.
    • Best practices for deploying and testing AI agents in cloud environments.
    • Community collaboration and shared learning resources.
  • Communication Preferences:
    • Technical documentation and tutorials that offer clear, concise instructions.
    • Community forums and discussions to share challenges and solutions.
    • Real-world use cases that demonstrate practical applications of new tools.

AWS Open-Sources an MCP Server for Bedrock AgentCore

AWS has released an open-source Model Context Protocol (MCP) server for Amazon Bedrock AgentCore, facilitating a direct transition from natural-language prompts to deployable agents. This solution includes automated transformations, environment provisioning, and tooling hooks, effectively compressing the typical multi-step integration process into conversational commands.

What is the AgentCore MCP Server?

The AgentCore MCP server provides task-specific tools to clients such as Kiro, Claude Code, Cursor, Amazon Q Developer CLI, and the VS Code Q plugin. It assists in:

  • Minimally refactoring existing agents for AgentCore Runtime model compatibility.
  • Provisioning and configuring the AWS environment, including credentials, roles, permissions, and configuration files.
  • Connecting the AgentCore Gateway for tool calls.
  • Invoking and testing the deployed agent directly from the IDE’s chat surface.

Essentially, the server guides the coding assistant to convert entry points to AgentCore handlers, manage imports, generate the requirements.txt, and execute direct agent calls through compatible payload-based handlers.

Installation and Client Support

Installing the MCP server is straightforward with a one-click install flow available from the GitHub repository utilizing a lightweight launcher (uvx) and a standard mcp.json entry. Various expected locations for mcp.json are:

  • Kiro: .kiro/settings/mcp.json
  • Cursor: .cursor/mcp.json
  • Amazon Q CLI: ~/.aws/amazonq/mcp.json
  • Claude Code: ~/.claude/mcp.json

The repository is hosted in the awslabs “mcp” mono-repo under an Apache-2.0 license, and it links to broader AWS MCP resources and documentation.

Architecture Guidance and Development Workflow

AWS recommends a layered approach for the IDE assistant to build progressively richer context:

  • Start with the agentic client.
  • Add the AWS Documentation MCP Server.
  • Incorporate framework documentation (Strands Agents, LangGraph).
  • Include AgentCore and agent-framework SDK docs.
  • Utilize per-IDE “steering files” for recurrent workflows.

This structured approach minimizes retrieval misses and enables the assistant to navigate the end-to-end process efficiently.

Typical Development Path

  • Bootstrap: Use local tools or MCP servers; provision a Lambda target for AgentCore Gateway or deploy directly to AgentCore Runtime.
  • Author/Refactor: Begin with Strands Agents or LangGraph code and use the server to modify handlers and dependencies for Runtime compliance.
  • Deploy: The assistant retrieves relevant documentation and invokes the AgentCore CLI for deployment.
  • Test & Iterate: Invoke the agent using natural language; if tool integrations are necessary, combine Gateway (MCP client inside the agent), redeploy (v2), and retest.

Impact of the MCP Server

The AgentCore MCP server simplifies the development workflow by shifting the burden of establishing cloud-specific knowledge onto the IDE assistant, thus narrowing the “prompt-to-production” gap. By integrating seamlessly with existing documentation servers (AWS service docs, Strands, LangGraph), it serves as a low-friction entry point for teams standardizing on Bedrock AgentCore.

Expert Insights

“The AWS MCP endpoint for AgentCore that my IDE can call directly makes the connection effortless, and the uvx-based mcp.json configuration streamlines client integration. The server’s tooling aligns well with the AgentCore Runtime and maintains compatibility with existing Strands and LangGraph code paths.”

Additional Resources

For further information on installation and usage, visit the AWS blog.

«`