lc2MCP (Xiaotonng)

lc2mcp

](https://pypi.org/project/lc2mcp/) ](https://pypi.org/project/lc2mcp/) 

将LangChain工具转换为FastMCP工具——只需一行代码。

停止重写你的工具。只是适应他们。

lc2mcp 是一个轻量级的适配器，可以转换现有的 LangChain工具 进入 FastMCP工具，使您能够快速构建可供Claude、Cursor和任何MCP兼容客户端访问的MCP服务器。

______________________________________________________________________

✨ 特性

______________________________________________________________________

🚀 快速开始

安装

pip install lc2mcp

通往MCP的3条线路

from langchain_core.tools import tool
from fastmcp import FastMCP
from lc2mcp import register_tools

@tool
def get_weather(city: str) -> str:
    """Get current weather for a city."""
    return f"Sunny, 25°C in {city}"

mcp = FastMCP("weather-server")
register_tools(mcp, [get_weather])  # ← That's it!

if __name__ == "__main__":
    mcp.run()

您的工具现在可供Claude、Cursor和任何MCP客户端使用。

刀具参数说明

要在MCP工具模式中包含参数描述，您有两个选项：

选项1：使用 parse_docstring=True （为简单起见，建议使用）

@tool(parse_docstring=True)
def get_weather(city: str, unit: str = "celsius") -> str:
    """Get current weather for a city.

    Args:
        city: The name of the city to query
        unit: Temperature unit (celsius or fahrenheit)
    """
    return f"Sunny, 25°C in {city}"

选项2：使用 args_schema （建议用于复杂类型）

from pydantic import BaseModel, Field

class WeatherInput(BaseModel):
    city: str = Field(description="The name of the city to query")
    unit: str = Field(default="celsius", description="Temperature unit")

@tool(args_schema=WeatherInput)
def get_weather(city: str, unit: str = "celsius") -> str:
    """Get current weather for a city."""
    return f"Sunny, 25°C in {city}"

注： 没有 parse_docstring=True 或 args_schema，将不会提取文档字符串中的参数描述。

______________________________________________________________________

🔌 运作原理

┌─────────────────┐      ┌─────────────┐      ┌─────────────────┐
│  LangChain Tool │ ───▶ │   lc2mcp    │ ───▶ │  FastMCP Tool   │
│  (@tool, etc.)  │      │  (adapter)  │      │                 │
└─────────────────┘      └─────────────┘      └────────┬────────┘
                                                       │
                                                       ▼
                                              ┌─────────────────┐
                                              │  FastMCP Server │
                                              └────────┬────────┘
                                                       │
                                                       ▼
                                          ┌───────────────────────┐
                                          │      MCP Clients      │
                                          │ (Claude, Cursor, ...) │
                                          └───────────────────────┘

______________________________________________________________________

🔄 lc2mcp与langchain mcp适配器

LangChain和MCP生态系统可以连接 双向:

何时使用 lc2mcp:

• 您有现有的LangChain工具，并希望通过MCP公开它们
• 您想使用LangChain丰富的工具生态系统构建MCP服务器
• 您需要为Claude、Cursor或其他MCP客户端提供工具

何时使用 langchain-mcp-adapters:

• 您有MCP服务器，并希望在LangChain代理中使用它们
• 您想从LangGraph工作流调用MCP工具

两者结合使用：

┌─────────────────────────────────────────────────────────────────┐
│                        Your Application                         │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│   LangChain Tools ──── lc2mcp ────▶ MCP Server ──▶ MCP Clients │
│         │                              │          (Claude, etc) │
│         │                              │                        │
│         ▼                              ▼                        │
│   LangChain Agent ◀── langchain-mcp-adapters ─── MCP Tools     │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

这两个库是互补的，可以一起用来构建连接LangChain和MCP生态系统的强大AI应用程序。

______________________________________________________________________

📚 例子

使用社区工具

立即将DuckDuckGo搜索和维基百科暴露给MCP客户端：

pip install lc2mcp langchain-community duckduckgo-search wikipedia

from fastmcp import FastMCP
from langchain_community.tools import DuckDuckGoSearchRun, WikipediaQueryRun
from langchain_community.utilities import WikipediaAPIWrapper
from lc2mcp import register_tools

mcp = FastMCP("knowledge-server")

register_tools(mcp, [
    DuckDuckGoSearchRun(),
    WikipediaQueryRun(api_wrapper=WikipediaAPIWrapper()),
])

if __name__ == "__main__":
    mcp.run()

使用身份验证上下文

将用户身份验证和应用程序上下文注入到您的工具中：

from dataclasses import dataclass
from fastmcp import Context, FastMCP
from langchain_core.tools import tool
from langgraph.prebuilt import ToolRuntime
from lc2mcp import register_tools

@dataclass(frozen=True)
class UserContext:
    user_id: str
    tenant_id: str

@tool
def whoami(runtime: ToolRuntime[UserContext]) -> str:
    """Return the current user."""
    return f"Hello, user {runtime.context.user_id} from {runtime.context.tenant_id}"

def runtime_adapter(mcp_ctx: Context) -> ToolRuntime[UserContext]:
    return ToolRuntime(
        context=UserContext(
            user_id=mcp_ctx.get_state("user_id") or "anonymous",
            tenant_id=mcp_ctx.get_state("tenant_id") or "default",
        ),
        state={}, config={}, stream_writer=lambda x: None,
        tool_call_id=None, store=None,
    )

mcp = FastMCP("auth-server")
register_tools(mcp, [whoami], runtime_adapter=runtime_adapter)

if __name__ == "__main__":
    mcp.run()

带有进度报告和记录功能

使用MCP上下文进行实时进度更新和记录：

from fastmcp import Context, FastMCP
from langchain_core.tools import tool
from lc2mcp import register_tools

@tool
async def process_data(data: str, mcp_ctx: Context) -> str:
    """Process data with progress reporting."""
    await mcp_ctx.info(f"Starting: {data}")
    await mcp_ctx.report_progress(0, 100, "Starting")
    
    # ... processing steps ...
    await mcp_ctx.report_progress(50, 100, "Processing")
    
    await mcp_ctx.info("Complete!")
    await mcp_ctx.report_progress(100, 100, "Done")
    return f"Processed: {data}"

mcp = FastMCP("processor")
register_tools(mcp, [process_data], inject_mcp_ctx=True)

if __name__ == "__main__":
    mcp.run()

命名空间和冲突处理

使用前缀组织工具并处理名称冲突：

from fastmcp import FastMCP
from lc2mcp import register_tools

mcp = FastMCP("multi-domain")

# Prefix all finance tools
register_tools(mcp, finance_tools, name_prefix="finance.")

# Auto-suffix on collision: tool → tool_2 → tool_3
register_tools(mcp, ops_tools, name_prefix="ops.", on_name_conflict="suffix")

if __name__ == "__main__":
    mcp.run()

______________________________________________________________________

📖 api参考

register_tools()

在服务器上将LangChain工具转换并注册为FastMCP工具。

register_tools(
    mcp: FastMCP,
    tools: list[BaseTool | Callable],
    *,
    name_prefix: str | None = None,           # e.g. "finance." → "finance.get_stock"
    on_name_conflict: str = "error",          # "error" | "overwrite" | "suffix"
    inject_mcp_ctx: bool = False,             # inject mcp_ctx: Context
    runtime_adapter: Callable | None = None,  # Context → ToolRuntime[...]
)

to_mcp_tool()

将单个LangChain工具转换为FastMCP工具进行手动注册。

to_mcp_tool(
    tool: BaseTool | Callable,
    *,
    name: str | None = None,
    description: str | None = None,
    args_schema: Type[BaseModel] | None = None,
    inject_mcp_ctx: bool = False,
    runtime_adapter: Callable | None = None,
) -> Callable

______________________________________________________________________

🔧 兼容性

工具支持

______________________________________________________________________

🤝 贡献

欢迎投稿！请随时提交拉取请求。

______________________________________________________________________

📄 许可证

MIT许可证-请参阅 许可证 了解详情。

______________________________________________________________________

Made with ❤️ for the LangChain and MCP communities