View a markdown version of this page

建立使用 AgentCore 閘道的代理程式 - Amazon Bedrock AgentCore

建立使用 AgentCore 閘道的代理程式

建立和測試閘道之後，您可以將 AI 代理器建立並連接至閘道。連接到閘道的代理程式可以呼叫閘道中的工具，並使用 Amazon Bedrock 模型來回應查詢。

若要了解如何建立代理程式、將其連接到閘道，並叫用它來回答查詢，請選取下列其中一種方法：

範例
Strands

  1. from strands import Agent
from strands.models import BedrockModel
from strands.tools.mcp.mcp_client import MCPClient
from mcp.client.streamable_http import streamablehttp_client

def _invoke_agent(
    bedrock_model,
    mcp_client,
    prompt
):
    with mcp_client:
        tools = mcp_client.list_tools_sync()
        agent = Agent(
            model=bedrock_model,
            tools=tools
        )
        return agent(prompt)

def _create_streamable_http_transport(headers=None):
    url = {gatewayUrl}
    access_token = {AccessToken}
    headers = {**headers} if headers else {}
    headers["Authorization"] = f"Bearer {access_token}"
    return streamablehttp_client(
        url,
        headers=headers
    )

def _get_bedrock_model(model_id):
    return BedrockModel(
        inference_profile_id=model_id,
        temperature=0.0,
        streaming=True,
   )

mcp_client = MCPClient(_create_streamable_http_transport)

if __name__ == "__main__":
    user_prompt = "What orders do I have?"
    _response = _invoke_agent(
        bedrock_model=_get_bedrock_model("us.anthropic.claude-sonnet-4-20250514-v1:0"),
        mcp_client=mcp_client,
        prompt=user_prompt
    )
    print(_response)
LangGraph

  1. from mcp import ClientSession
from mcp.client.streamable_http import streamablehttp_client

from langgraph.prebuilt import create_react_agent
from langchain_mcp_adapters.tools import load_mcp_tools

# Replace with actual values and the Amazon Bedrock model of your choice
gateway_url = "${GatewayUrl}"
access_token = "${AccessToken}"
model = ChatBedrock(model_id="anthropic.claude-3-sonnet-20240229-v1:0", region_name="us-west-2")

async with streamablehttp_client(gateway_url, headers={"Authorization": f"Bearer {access_token}"}) as (read, write, _):
    async with ClientSession(read, write) as session:
        # Initialize the connection
        await session.initialize()

        # Get tools
        tools = await load_mcp_tools(session)
        agent = create_react_agent(model, tools)
        math_response = await agent.ainvoke({"messages": "what's (3 + 5) x 12?"})
Claude Code

  1. #!/bin/bash
# Script to add MCP server to Claude

# Server configuration
SERVER_NAME=${ServerName} # Write your server name
GATEWAY_MCP_SERVER_URL=${GatewayUrl} # The gateway MCP URL
AUTH_TOKEN=${AuthToken} # Claude authentication token

echo "Adding MCP server to Claude..."
echo "Server Name: $SERVER_NAME"
echo "Server URL: $SERVER_URL"
echo ""

# Add the MCP server
claude mcp add "$SERVER_NAME" "$GATEWAY_MCP_SERVER_URL" \
  --transport http \
  --header "Authorization: Bearer $AUTH_TOKEN"

# Check if the command was successful
if [ $? -eq 0 ]; then
    echo "MCP server added successfully!"
    echo ""
    echo "You can now check mcp server health with: claude mcp list"
else
    echo "Failed to add MCP server"
    exit 1
fi