Google has open-sourced a Model Context Protocol (MCP) server for its Ads API, giving developers a standardized way to let AI agents analyze advertising data using natural language, as first reported by Search Engine Land. The move simplifies building AI-powered marketing tools and helps establish MCP as an industry-wide connector for AI and external apps.
Look but don't touch: The experimental server operates in a strictly read-only mode, a safeguard to prevent AI from accidentally altering campaigns or running up an ad bill. It lets developers use AI to search for campaign data with Google's query language or get a list of accessible customer accounts, processing natural language requests for performance metrics.
A new standard emerges: The move is more than just a new developer tool; it helps cement MCP's role as an industry-wide connector for AI and SaaS platforms. It makes it dramatically easier for marketing and growth teams to experiment with agentic workflows for tasks like performance diagnostics and automated reporting, all without handing over write-access.
While the tool is read-only for now, it previews a future where AI agents could expand beyond simply reporting on campaigns to directly managing and optimizing them.
Google is also making its developer tools more AI-native with a new Genkit extension for the Gemini CLI. The release comes as the industry pushes toward more advanced "computer-use agents" that can operate software without APIs, and new research explores "agentic context engineering" as a way to make AI models smarter without retraining them.
