Dolphin MCP: Cross-Provider Tools

This technical deep-dive demonstrates how Dolphin MCP enables seamless interaction with multiple LLM providers (OpenAI, Anthropic, Ollama, LMStudio) through a unified Model Context Protocol (MCP) architecture. The presentation will feature:

Live coding session showing how to build a modular AI application that connects to multiple model providers simultaneously
Technical walkthrough of the MCP client implementation, exploring function calling, server discovery, and asynchronous tool execution
Interactive demonstration of real-time tool orchestration across different LLM backends, with audience members suggesting queries
Code dissection of key components including provider-specific implementations, showing how to handle the differences between OpenAI, Anthropic, Ollama, and LMStudio APIs
Architecture analysis of the project’s approach to separating concerns between providers, tools, and client interfaces

I’ll show concrete examples of how Dolphin MCP handles function calling across different providers that implement it differently, including explicit handling for:

OpenAI’s native function calling
Anthropic’s tool use format
Ollama’s function calling implementation
LMStudio’s unique approach

The session will include debugging challenging edge cases in multi-provider environments and demonstrating how to build resilient systems that work even when providers return unexpected responses.