与代理框架一起使用

您可以将 OpenSearch MCP 服务器直接集成到 Python 代理框架中，从而在更大的工作流程中 OpenSearch 为自治代理提供编程访问权限。

Strands Agents

Strands Agents 是一款 AWS原生代理 SDK，内置 MCP 支持，Amazon Bedrock 是默认模型提供商。以下示例通过设置适当的环境变量将 Strands 代理连接到 OpenSearch 服务域或 OpenSearch 无服务器集合。连接AWS_OPENSEARCH_SERVERLESS到 OpenSearch Serverless 集合true时设置为；对于托管域，则将其省略。

import os
from strands import Agent
from strands.tools.mcp import MCPClient
from mcp import stdio_client, StdioServerParameters

# For a managed domain:
#   OPENSEARCH_URL = https://<domain-endpoint>.<region>.es.amazonaws.com
#
# For an OpenSearch Serverless collection, also set:
#   AWS_OPENSEARCH_SERVERLESS = true
#   OPENSEARCH_URL = https://<collection-id>.<region>.aoss.amazonaws.com
opensearch_client = MCPClient(
    lambda: stdio_client(
        StdioServerParameters(
            command="uvx",
            args=["opensearch-mcp-server-py"],
            env={
                "OPENSEARCH_URL":            os.environ["OPENSEARCH_URL"],
                "AWS_REGION":                os.environ["AWS_REGION"],
                "AWS_IAM_ARN":               os.environ["AWS_IAM_ARN"],
                # Set to "true" for OpenSearch Serverless, omit for managed domains
                "AWS_OPENSEARCH_SERVERLESS": os.environ.get("AWS_OPENSEARCH_SERVERLESS", "false"),
            },
        )
    )
)

with opensearch_client:
    agent = Agent(tools=opensearch_client.list_tools_sync())
    response = agent("List all indexes and show the document count for each")
    print(response)

Strands 使用 Amazon Bedrock 作为其默认模型提供商。确保在您所在地区为 Claude 配置了 AWS 凭证并启用了模型访问权限。有关详细信息，请参阅 Strands Bedrock 提供商文档。

LangGraph

LangGraph是用于构建有状态代理的低级编排框架。以下示例用于将 OpenSearch MCP 工具加载langchain-mcp-adapters到由 Amazon Bedrock 支持的 LangGraph ReAct 代理中。与 Strands 一样，true在AWS_OPENSEARCH_SERVERLESS连接到 OpenSearch 无服务器集合时将其设置为。

import asyncio
import os
from langchain_aws import ChatBedrock
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent

async def main():
    async with MultiServerMCPClient(
        {
            "opensearch": {
                "command": "uvx",
                "args": ["opensearch-mcp-server-py"],
                "env": {
                    # Managed domain:  https://<domain-endpoint>.<region>.es.amazonaws.com
                    # Serverless:      https://<collection-id>.<region>.aoss.amazonaws.com
                    "OPENSEARCH_URL":            os.environ["OPENSEARCH_URL"],
                    "AWS_REGION":                os.environ["AWS_REGION"],
                    "AWS_IAM_ARN":               os.environ["AWS_IAM_ARN"],
                    # Set to "true" for OpenSearch Serverless, omit for managed domains
                    "AWS_OPENSEARCH_SERVERLESS": os.environ.get("AWS_OPENSEARCH_SERVERLESS", "false"),
                },
                "transport": "stdio",
            }
        }
    ) as mcp_client:
        tools = mcp_client.get_tools()
        model = ChatBedrock(
            model_id="anthropic.claude-3-5-sonnet-20241022-v2:0",
            region_name=os.environ["AWS_REGION"],
        )
        agent = create_react_agent(model, tools)
        result = await agent.ainvoke(
            {"messages": [{"role": "user", "content": "Check cluster health and list all indexes"}]}
        )
        print(result["messages"][-1].content)

asyncio.run(main())

安装所需的软件包：

pip install langchain-aws langchain-mcp-adapters langgraph