记忆-MCP-焦点-LangChain（或“记忆-多模态控制协议-焦点-语言链”）
](https://badge.fury.io/py/mimory-mcp-focus-langchain)  
LangChain 的集成用于 “mimory-mcp-focus”可以翻译为“记忆-MCP（多通道处理/模块控制平台）焦点”或根据具体上下文调整为更贴切的表述，如“记忆-MCP核心关注点”等。这里，“mimory”可能是一个特定于某个领域或产品的术语，我将其翻译为“记忆”，但具体含义可能需要根据上下文来确定；“mcp”通常代表“Multi-Channel Processing”（多通道处理）或“Module Control Platform”（模块控制平台）等，具体含义同样取决于上下文；“focus”则翻译为“焦点”或“核心关注点”为您的LangChain MCP实现提供焦点（或核心）功能。
概述
mimory-mcp-focus-langchain 扩展了LangChain的MCP（模型上下文协议）集成，增强了其聚焦功能。它使您能够：
- 限制工具访问 针对特定的MCP工具
- 控制参数值 带有白名单和范围
- 应用复合焦点 对于复杂的工具特定限制
- 无缝集成 借助LangChain的工具生态系统
安装
pip install mimory-mcp-focus-langchain依赖项
这个软件包需要：
- Python 3.10及以上版本
- 记忆-MCP焦点 - 核心关注功能
- LangChain-MCP适配器 - LangChain与MCP集成
- LangChain-Core（或“语言链核心”） - LangChain 核心功能
快速入门
使用 FocusMultiServerMCPClient 的基本用法
import asyncio
from mim_mcp_focus_langchain import FocusMultiServerMCPClient
from langchain_mcp_adapters.sessions import Connection
async def main():
# Define MCP server connections
connections = {
"everything": {
"transport": "stdio",
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-everything"]
},
"time": {
"transport": "stdio",
"command": "python",
"args": ["-m", "mcp_server_time"]
}
}
# Configure focus restrictions
focus_configs = {
"everything": {
"focus_tools": ["echo", "add"], # Only allow these tools
"focus_params": {
"message": ["hello", "world"], # Restrict message values
"a": ["range:1-10"] # Restrict 'a' parameter to 1-10
}
},
"time": {
"focus_tools": ["get_current_time"], # Only allow time tool
"focus_params": {
"timezone": ["America/New_York", "Europe/London"] # Restrict timezones
}
}
}
# Create focus-enabled client
client = FocusMultiServerMCPClient(connections, focus_configs)
# Get tools with focus applied
tools = await client.get_tools()
# Use tools in your LangChain application
for tool in tools:
print(f"Tool: {tool.name} - {tool.description}")
if __name__ == "__main__":
asyncio.run(main())使用LangChain结合Core FocusClient
import asyncio
from mimory_mcp_focus import focus_client_simple, focus_client_composite
from langchain_mcp_adapters.sessions import Connection, create_session
from langchain_mcp_adapters.tools import load_mcp_tools
async def main():
# Create MCP connection
connection = {
"transport": "stdio",
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-everything"]
}
# Create focus-enabled session
async with create_session(connection) as session:
await session.initialize()
# Apply focus restrictions using simple focus
focus_session = focus_client_simple(
session,
focus_tools=["echo", "add"],
focus_params={
"message": ["hello", "world"],
"a": ["range:1-10"]
},
strict=False
)
# Load tools with focus applied
tools = await load_mcp_tools(focus_session)
# Use tools in LangChain
for tool in tools:
print(f"Tool: {tool.name}")
if __name__ == "__main__":
asyncio.run(main())焦点配置
简单聚焦
全局限制工具和参数：
focus_config = {
"focus_tools": ["tool1", "tool2"], # Allowed tools
"focus_params": {
"param1": ["value1", "value2"], # Allowed values
"param2": ["range:1-100"] # Numeric range
},
"focus_type": "simple", # How the focus params are encoded
"strict": False # Whether to enforce strict mode
}复合焦点
为每种工具应用不同的限制：
focus_config = {
"focus": {
"tool1": {
"param1": ["value1", "value2"],
"param2": ["range:1-10"]
},
"tool2": {
"param3": ["value3", "value4"],
"param4": ["value5"]
}
},
"focus_type": "composite", # How the focus params are encoded
"strict": True
}焦点配置结构
该 focus_config 字典支持以下结构：
对于简易对焦（focus_type: "simple"):
focus_tools允许使用的工具名称列表（使用["*"]允许所有工具）focus_params参数名称到允许值的字典映射focus_type必须是"simple"strict布尔值，指示是否启用严格模式
对于复合焦点（focus_type: "composite"):
focus字典，用于将工具名称映射到其参数限制focus_type必须是"composite"strict布尔值，指示是否启用严格模式
参数值格式：
- 确切数值：
["value1", "value2"] - 数字范围：
["range:1-100"]（包括） - 通配符：
["*"]（允许任何值）
API 参考文档
FocusMultiServerMCPClient 翻译为中文是：“焦点多服务器MCP客户端”
用于支持焦点功能的多服务器MCP操作的主类。
构造函数
FocusMultiServerMCPClient(
connections: Optional[Dict[str, Connection]] = None,
focus_configs: Optional[Dict[str, Dict[str, Any]]] = None
)参数：
connections将服务器名称映射到连接配置的字典focus_configs字典，用于将服务器名称映射到焦点配置
方法
async get_tools(server_name: Optional[str] = None) -> List[BaseTool]
获取应用了焦点限制的LangChain工具。
参数：
server_name可选的特定服务器名称。如果为 None，则返回所有服务器的工具。
返回：
- 应用了重点的LangChain工具列表
async session(server_name: str, *, auto_initialize: bool = True) -> AsyncIterator[ClientSession]
创建一个启用焦点功能的MCP会话。
参数：
server_name要连接的服务器名称auto_initialize是否自动初始化会话
返回值：
- 异步上下文管理器，生成一个启用焦点的ClientSession
示例
示例1：基本工具限制
# examples/basic_restriction.py
import asyncio
from mim_mcp_focus_langchain import FocusMultiServerMCPClient
async def main():
connections = {
"everything": {
"transport": "stdio",
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-everything"]
}
}
focus_configs = {
"everything": {
"focus_tools": ["echo"], # Only allow echo tool
"focus_params": {
"message": ["hello", "world", "test"] # Restrict message values
},
"focus_type": "simple",
"strict": False
}
}
client = FocusMultiServerMCPClient(connections, focus_configs)
tools = await client.get_tools()
print("Available tools:")
for tool in tools:
print(f"- {tool.name}: {tool.description}")
if __name__ == "__main__":
asyncio.run(main())示例2：具有复合焦点的多服务器
# examples/multi_server_composite.py
import asyncio
from mim_mcp_focus_langchain import FocusMultiServerMCPClient
async def main():
connections = {
"everything": {
"transport": "stdio",
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-everything"]
},
"time": {
"transport": "stdio",
"command": "python",
"args": ["-m", "mcp_server_time"]
}
}
focus_configs = {
"everything": {
"focus": {
"echo": {
"message": ["hello", "world"]
},
"add": {
"a": ["range:1-5"],
"b": ["range:1-5"]
}
},
"focus_type": "composite",
"strict": False
},
"time": {
"focus_tools": ["get_current_time"],
"focus_params": {
"timezone": ["America/New_York", "Europe/London"]
},
"focus_type": "simple",
"strict": False
}
}
client = FocusMultiServerMCPClient(connections, focus_configs)
tools = await client.get_tools()
print("Available tools:")
for tool in tools:
print(f"- {tool.name}: {tool.description}")
if __name__ == "__main__":
asyncio.run(main())示例3：与LangChain代理一起使用
# examples/langchain_agent.py
import asyncio
from mim_mcp_focus_langchain import FocusMultiServerMCPClient
from langchain.agents import AgentExecutor, create_tool_calling_agent
from langchain_openai import ChatOpenAI
async def main():
# Setup MCP connections with focus
connections = {
"everything": {
"transport": "stdio",
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-everything"]
}
}
focus_configs = {
"everything": {
"focus_tools": ["echo", "add"],
"focus_params": {
"message": ["hello", "world", "test"],
"a": ["range:1-10"],
"b": ["range:1-10"]
},
"focus_type": "simple",
"strict": False
}
}
# Create focus-enabled client
client = FocusMultiServerMCPClient(connections, focus_configs)
# Get tools
tools = await client.get_tools()
# Create LangChain agent
llm = ChatOpenAI(model="gpt-4", temperature=0)
agent = create_tool_calling_agent(llm, tools, "You are a helpful assistant with access to focused MCP tools.")
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
# Run agent
result = await agent_executor.ainvoke({"input": "Echo 'hello' and add 5 + 3"})
print(result)
if __name__ == "__main__":
asyncio.run(main())错误处理
该库使用标准的Python异常来处理错误。与焦点相关的错误将作为适当的异常抛出，您可以根据需要捕获并处理这些异常。
做出贡献
欢迎贡献！请随时提交拉取请求。
许可证
这个项目采用MIT许可证授权——详见 许可证 详情请见文件。
链接
- LangChain Python SDK（软件开发工具包）
- LangChain MCP 适配器
- mimory-mcp-focus 翻译为中文可以是“记忆-MCP（多通道处理）焦点”或“记忆-多通道处理关注点”，具体翻译取决于上下文和应用场景。在这里，“mimory”可能是一个特定领域或产品中的术语，假设它与“记忆”相关；“MCP”通常代表“Multi-Channel Processing”（多通道处理）；“focus”则意为“焦点”或“关注点”。因此，整个短语可以理解为与记忆相关的多通道处理中的焦点或关注点
- 模型上下文协议
作者
特拉维斯·麦克奎因 - oss@mimory.io（注：这是一封电子邮件地址，直接翻译为中文形式即保持原样，因为电子邮件地址在中文中通常也直接使用其原始形式。）
支持
如需支持，请在（相关平台/项目）上提交一个问题 。