>But to your LLM, there’s no difference between “regular” tool calling and using a standard like MCP.
The difference is that it's trained and benchmarked on the MCP-standard. Which makes it much more reliable than if you create your own interface.
Extremely insightful and detailed. Thank you! Could you create a concise YouTube explainer on this?
Also, what are the best strategies to rigorously validate inputs while minimizing latency?
Is this the best for Rust: https://github.com/modelcontextprotocol/rust-sdk