MCP Server Hub

MCP Server Hub is the unified service for deploying, configuring, and monitoring MCP servers on AgentOpera Developer Platform. Its primary purpose is to simplify the deployment of various MCP servers, accelerate AI Agent integration with third-party tools, and provide developers with a one-stop MCP service management experience.

To learn more about MCP details, please refer to the official documentation.

1. MCP Hub Architecture

MCP Hub builds a backend service that acts as a proxy between MCP clients and various MCP servers. The system is designed as a bridge to connect AI applications with tool providers while abstracting away the complexities of different protocols and implementation details.

The architecture diagram illustrates the three-layered design of MCP Hub:

Client Applications Layer

Various clients (AgentOpera Framework, Low-code AI Tools, Custom MCP Clients) consume MCP services through a standardized HTTP/SSE interface.

MCP Hub Layer

The core platform with several key components that work together:

  • Proxy Server: Handles requests from MCP clients and routes them to the appropriate MCP server

  • SSE Sessions Management: Manages real-time communication between clients and MCP servers

  • Client Adapters: Handles communication with different types of MCP servers

MCP Servers Layer

The actual tool implementations, including heterogeneous MCP servers built with different technologies (uvx/npx/python) and pre-configured services with standard transport protocols.

2. Core Capabilities

  • Full Lifecycle Management: Provides comprehensive lifecycle management capabilities including deployment, initialization configuration, version control, and operational monitoring, helping MCP services run stably and efficiently.

  • Unified Control Plane: Focuses on unified orchestration and operational control of different MCP Servers, building a centralized management architecture to achieve consistent management across environments and platforms.

  • Heterogeneous Protocol Abstraction: Shields the underlying heterogeneous transport protocol differences between different MCP Servers, provides standardized unified API access, and simplifies the integration process for Agent framework MCP Clients.

  • One-Click MCP Service Enablement: Users can quickly enable common MCP capabilities through one-click activation, shortening the deployment cycle of AI capabilities.

3. Platform Interface

MCP Servers Marketplace

The MCP Servers Marketplace showcases platform-provided and developer-submitted MCP Servers. Developers can browse, search, and select MCP Servers suitable for their application scenarios.

MCP Server Connection Interface

Through the MCP Server connection interface, users can input necessary parameters (such as API KEY) to start and configure business MCP Servers. After a successful connection, users can view the MCP Server's operational status.

MCP Server Details Interface

The MCP Server details interface displays detailed information about the server, including:

  • Server introduction and functional description

  • List of supported tools and usage examples

  • SSE configuration guide (applicable for different environments like Claude, Cursor, etc.)

4. Integrating MCP in AgentOpera

AgentOpera provides simple and easy-to-use APIs that enable developers to easily integrate MCP services. Below is the basic integration process:

import asyncio
from agentopera.models.openai import OpenAIChatCompletionClient
from agentopera.agents.tools.mcp import SseServerParams, mcp_server_tools
from agentopera.chatflow.agents import AssistantAgent
from agentopera.core import CancellationToken

async def main() -> None:
    # Set up MCP server connection parameters
    server_params = SseServerParams(
        url="https://open.tensoropera.ai/mcp-hub/sse/1d08ff70xxxxyyyy", 
    )
    
    # Get all tools provided by the server
    tools = await mcp_server_tools(server_params)
    
    # Create an Agent that can use these tools
    agent = AssistantAgent(
        name="tool_user",
        model_client=OpenAIChatCompletionClient(model="gpt-4"),
        tools=tools  # type: ignore
    )
    
    # Use the Agent to complete tasks
    result = await agent.run(
        task="Use tools to complete specific tasks", 
        cancellation_token=CancellationToken()
    )
    
    print(result.messages[-1].content)

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

5. Integrating MCP through Low-code AI Tools

In addition to code integration, MCP Hub also supports integrating MCP services through Low-code AI Tools at AgentOpera Developer Platform.

In the flow editor, users can:

  1. Select an MCP connection node

  2. Configure the connection protocol (typically SSE protocol)

  3. Input the SSE URL and necessary authentication information

  4. Connect the MCP node with other components (such as LLM, input/output nodes, etc.) to build a complete application flow

PreviousStream Protocol with WebUINextIntegrations

Last updated