mcp-client-demo

# MCP Client ## Contents - Implement a basic Client to integrate deepseek into the MCP ecosystem - Core Architecture - JSON-RPC 2.0 - What is RPC? - **Differences between MCP and Function Calling** - Differences in APIs among various providers - OpenAI / Anthropic / DeepSeek - **Refer to the 5ire source code for smoothing out API differences** ## MCP Architecture https://modelcontextprotocol.io/docs/concepts/architecture https://www.jsonrpc.org/ ## Implementation of MCP Client https://modelcontextprotocol.io/quickstart/client https://github.com/modelcontextprotocol/typescript-sdk - MCP Server: https://github.com/pskill9/hn-server - Inspector: https://github.com/modelcontextprotocol/inspector ## Differences between MCP and Function Calling ### Tool Invocation in MCP ```mermaid graph LR LLM <--> MCP_Client <--> MCP_Server ``` - There is a strict communication protocol between MCP Client and MCP Server - The server provides a unified interface specification and service registration mechanism - The client knows how to communicate with the server and the available service list - Due to the **unified protocol specification**, tools developed by different developers can be easily reused by other MCP clients - The server's registration mechanism simplifies tool discovery and sharing ### Tool Invocation in Function Calling ```mermaid graph LR LLM <--> AnyClient <--> AnyServer ``` - It is more like a **basic capability** of the model, where the LLM is only responsible for identifying what functionality needs to be called - The client decides how to implement this function call - There is no unified service discovery and communication standard - The plugin ecosystems of different clients cannot be compatible with each other - https://chat-plugin-sdk.lobehub.com/quick-start/get-start - https://www.librechat.ai/docs/development/tools_and_plugins ## API Differences https://platform.openai.com/docs/api-reference/chat https://docs.anthropic.com/en/api/messages https://api-docs.deepseek.com/api/create-chat-completion - Differences 1. The function definition format of function_calling 2. The data structure returned by function_calling 3. The data structure of the message after the call ends ## How 5ire Smooths Out API Differences - The API return data structures of various models are different; how to smooth out the differences between LLM model outputs and mcp-client? - Since we know they are different → we can convert them to be the same - https://github.com/nanbingxyz/5ire 1. Different function definition methods → handled by ChatService's makeTool 2. Different function call results → handled by Reader's parseReply and parseTools 3. Different message formats → handled by ChatService's makeToolMessages