Google Cloud Next 25 brings A2A and conversational analytics in preview
At Next 25, Google announced the Agent2Agent protocol as a draft open standard and put conversational analytics for Looker and a new API into preview, both grounded in the Looker semantic model.
Google Cloud Next ran this week in Las Vegas. Two of its announcements bear directly on agents over business data.
Agent2Agent, a draft protocol
Google announced Agent2Agent (A2A), an open protocol for agents to find each other and hand off work. It launched as a draft specification with more than 50 partners, and Google positioned it as a complement to MCP: MCP connects an agent to tools and data, and A2A connects agents to other agents. Google said a production version is planned for later this year.
Conversational analytics, in preview
- Looker conversational analytics lets people ask questions of their Looker data in plain language. It is in preview.
- The Conversational Analytics API, also in preview, lets developers build the same kind of question answering into their own apps.
- BigQuery knowledge engine, in preview, uses Gemini to read schema relationships and query history, then generates metadata and suggests business glossary terms.
- MCP Toolbox for Databases launched, and Google's Agent Development Kit supports MCP.
Why it matters
Google grounds these answers in the Looker semantic model, the place where a company has already written down what its measures mean. That is the right instinct. A model that guesses what "revenue" means from column names will guess differently than your finance team does.
The open question for most companies is the part a semantic model does not cover. It defines measures and joins. It does not usually say what a customer is, which events change a customer's status, or which team's definition wins when two disagree. Agents will need that too.