home / mcp / aws agentcore mcp server

AWS AgentCore MCP Server

Provides MCP access to AWS AgentCore components for building secure, observable AI agents at scale.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "weiwarren-agentcore-mcp-server": {
      "command": "uvx",
      "args": [
        "aws-agentcore-mcp-server"
      ]
    }
  }
}

You can connect to and run the AWS AgentCore MCP Server to build, secure, and observe AI agents at scale. This server provides a structured way to manage agent identities, gateway integrations, observability, secure code execution, and memory, all through MCP-compatible tooling.

How to use

You will use MCP clients to connect to the AWS AgentCore MCP Server and issue requests that leverage AgentCore Identity, Gateway, Observability, Code Interpreter, and Memory features. Start by ensuring the MCP client is configured to reach the local or remote MCP endpoint exposed by uvx. Once connected, you can explore capabilities such as secure agent authentication, API integration, production observability, sandboxed code execution, and memory-backed context for your GenAI workflows.

How to install

Prerequisites: ensure uv is installed on your system. Follow the official installation instructions for uv before proceeding.

Install and run the AWS AgentCore MCP Server locally using the standard MCP runtime. The server is invoked via uvx with the target MCP server binary or script named aws-agentcore-mcp-server.

Step-by-step commands you can run to prepare and start the MCP server locally:

# Example: prepare environment and start the MCP server locally
# 1) Ensure uv is installed per your platform
# 2) Run the MCP server using the runtime wrapper
uvx aws-agentcore-mcp-server

Additional configuration notes

The following MCP configuration example shows how to register the AWS AgentCore MCP Server under a standard MCP runtime configuration. This entry uses a local stdio session to start the server process controlling aws-agentcore-mcp-server.

{
  "mcpServers": {
    "aws_agentcore": {
      "command": "uvx",
      "args": ["aws-agentcore-mcp-server"]
    }
  }
}

Usage patterns with common MCP clients

If you use the Q Developer CLI, Claude Code, Cline, or Cursor to manage MCP configurations, you can add the AWS AgentCore MCP Server using the corresponding commands shown in typical setup flows. Each method is designed to register the same stdio-based server connection described above, ensuring your MCP client can start, monitor, and interact with AgentCore components.

Available tools

quickstart

Get started with AWS AgentCore SDK and MCP client onboarding.

agentcore_identity

Learn about secure agent authentication and how to grant and verify credentials.

agentcore_gateway

Integrate external APIs and services through the AgentCore Gateway.

agentcore_observability

Monitor, trace, and debug agents in production with observability features.

agentcore_code_interpreter

Execute code securely within sandboxed sessions for agents.

agentcore_memory

Enable memory mechanisms for context-aware agent behavior, including short-term and long-term memory.

agentcore_tools

Extend agent capabilities by integrating tools and invoking external utilities.