Even the best AI agents are thwarted by this protocol - what can be done

"An emerging category of artificial intelligence middleware known as Model Context Protocol is meant to make generative AI programs such as chatbots bots more powerful by letting them connect with various resources, including packaged software such as databases. Multiple studies, however, reveal that even the best AI models struggle to use Model Context Protocol. Top AI models such as Google's Gemini 5 require many, many rounds of interactions with the external programs, leading to long delays in the performance of the AI models."

""Performance generally declines as tasks transition from Single Server to Multi Server scopes," writes Zikang Guo and team at the University of Science and Technology of China last month when they tested several AI models on their own benchmark test, MCP-AgentBench. Even the best models today, including OpenAI's GPT-5, have "failure cases" arising from "repetitive or exploratory interactions that fail to make meaningful progress,""

Model Context Protocol (MCP) is middleware that enables generative AI to connect to external resources and packaged software such as databases. Multiple benchmark suites show top models struggle to use MCP effectively, particularly as tasks scale from single-server to multi-server environments. State-of-the-art models like Gemini 5 and GPT-5 often require many interaction rounds with external programs, causing long delays and repetitive exploratory actions that make little progress. Performance generally declines as task complexity increases. Additional, MCP-specific training and development of model capabilities are required to reduce failures and improve efficiency.

Read at ZDNET