HarmonyOS MCP Server

概要

HarmonyOS デバイスを操作するための MCP サーバーです。

https://github.com/user-attachments/assets/7af7f5af-e8c6-4845-8d92-cd0ab30bfe17

クイックスタート

インストール

1. このリポジトリをクローンします。

git clone https://github.com/XixianLiang/HarmonyOS-mcp-server.git
cd HarmonyOS-mcp-server

2. 環境をセットアップします。

uv python install 3.13
uv sync

使い方

1.Claude Desktop

Claude Desktop を使って、このツールをお試しいただけます。

2.Openai SDK

openai-agents SDK を使って、MCP サーバーをお試しいただくこともできます。以下に例を示します。

"""
Example: Use Openai-agents SDK to call HarmonyOS-mcp-server
"""
import asyncio
import os

from agents import Agent, Runner, gen_trace_id, trace
from agents.mcp import MCPServerStdio, MCPServer

async def run(mcp_server: MCPServer):
    agent = Agent(
        name="Assistant",
        instructions="Use the tools to manipulate the HarmonyOS device and finish the task.",
        mcp_servers=[mcp_server],
    )

    message = "Launch the app `settings` on the phone"
    print(f"Running: {message}")
    result = await Runner.run(starting_agent=agent, input=message)
    print(result.final_output)


async def main():

    # Use async context manager to initialize the server
    async with MCPServerStdio(
        params={
            "command": "<...>/bin/uv",
            "args": [
                "--directory",
                "<...>/harmonyos-mcp-server",
                "run",
                "server.py"
            ]
        }
    ) as server:
        trace_id = gen_trace_id()
        with trace(workflow_name="MCP HarmonyOS", trace_id=trace_id):
            print(f"View trace: https://platform.openai.com/traces/trace?trace_id={trace_id}\n")
            await run(server)

if __name__ == "__main__":
    asyncio.run(main())

3.Langchain

柔軟な LLM エージェントフレームワークである LangGraph を使って、ワークフローを設計できます。以下に例を示します。

"""
langgraph_mcp.py
"""

server_params = StdioServerParameters(
    command="/home/chad/.local/bin/uv",
    args=["--directory",
          ".",
          "run",
          "server.py"],
    
)


#This fucntion would use langgraph to build your own agent workflow
async def create_graph(session):
    llm = ChatOllama(model="qwen2.5:7b", temperature=0)
    #!!!load_mcp_tools is a langchain package function that integrates the mcp into langchain.
    #!!!bind_tools fuction enable your llm to access your mcp tools
    tools = await load_mcp_tools(session)
    llm_with_tool = llm.bind_tools(tools)

    
    system_prompt = await load_mcp_prompt(session, "system_prompt")
    prompt_template = ChatPromptTemplate.from_messages([
        ("system", system_prompt[0].content),
        MessagesPlaceholder("messages")
    ])
    chat_llm = prompt_template | llm_with_tool

    # State Management
    class State(TypedDict):
        messages: Annotated[List[AnyMessage], add_messages]

    # Nodes
    def chat_node(state: State) -> State:
        state["messages"] = chat_llm.invoke({"messages": state["messages"]})
        return state

    # Building the graph
    # graph is like a workflow of your agent.
    #If you want to know more langgraph basic,reference this link (https://langchain-ai.github.io/langgraph/tutorials/get-started/1-build-basic-chatbot/#3-add-a-node)
    graph_builder = StateGraph(State)
    graph_builder.add_node("chat_node", chat_node)
    graph_builder.add_node("tool_node", ToolNode(tools=tools))
    graph_builder.add_edge(START, "chat_node")
    graph_builder.add_conditional_edges("chat_node", tools_condition, {"tools": "tool_node", "__end__": END})
    graph_builder.add_edge("tool_node", "chat_node")
    graph = graph_builder.compile(checkpointer=MemorySaver())
    return graph





async def main():
    async with stdio_client(server_params) as (read, write):
        async with ClientSession(read, write) as session:
            await session.initialize()

            config = RunnableConfig(thread_id=1234,recursion_limit=15)
            # Use the MCP Server in the graph
            agent = await create_graph(session)

            while True:
                message = input("User: ")
                try:
                    response = await agent.ainvoke({"messages": message}, config=config)
                    print("AI: "+response["messages"][-1].content)
                except RecursionError:
                    result = None
                    logging.error("Graph recursion limit reached.")


if __name__ == "__main__":
    asyncio.run(main())

server.py にシステムプロンプトを記述します。

"""
server.py
"""
@mcp.prompt()
def system_prompt() -> str:
    """System prompt description"""
    return """
    You are an AI assistant use the tools if needed.
    """

load_mcp_prompt 関数を使って、MCP サーバーからプロンプトを取得します。

"""
langgraph_mcp.py
"""
prompts = await load_mcp_prompt(session, "system_prompt")