Instead of forcing coding agents to invent non-existent methods or use outdated library syntax, Google implemented support for the Model Context Protocol (MCP). Now the agent dynamically pulls current documentation into the context before writing code. Google claims a pass rate of 96.3% on its internal tests. Ironically, the MCP standard was originally developed by competitors from Anthropic. The fact that Google is adapting it says one thing: the industry is finally forming unified standards for agent interaction with the outside world, turning AI from a text generator into a reliable engineer.
Source: Google DeepMind / Tech Buzz
DevelopmentGoogleGeminiMCPAPI