Google is changing its huge public data into a goldmine for AI with the introduction of the data common model reference protocol (MCP) Server-enabling the natural language and a better train AI system to enable AI to AI for AI-Developers, data scientists and AI agents.
Launched in 2018, Google’s data Commons organizes public dataset from one Boundary of sourcesGovernment surveys, local administrative data and data from global bodies such as the United Nations include. With the release of the MCP server, this data is now accessible through the natural language, allowing developers to integrate it in AI agents or applications.
The AI system is often trained on noise, rejected web data. When there is a shortage of sources, it is combined with their tendency to “fill the spaces”, this causes hallucinations. As a result, the Fine-tuned companies often require access to large, high quality datasets for specific use cases. By releasing the MCP server for publicly publicly, Google aims to deal with both challenges.
The new MCP server of data commons brids public dataset – from census data to climate data – with the AI system that depend on the rapidly accurate, structured context. By making this data accessible through natural language signals, the purpose of release is to keep AI on the ground in real world knowledge.
“Model Reference Protocol is using the intelligence of big language models to choose the right data at the right time,” the head of Google Data Commons, Prem Ramaswamy said in an interview, without understanding how we model the data, how our API works. “
Previously introduced by anthropic in last November, MCP is an open industry standard that enables the AI system to access data from various sources, including business equipment, material repository and app development environments, providing a common framework to understand the relevant signals. Since its launch, companies like Openai, Microsoft, and Google The standard is adopted to integrate your AI model with various data sources.
While other technical companies discovered how to implement the standard in their AI model, Ramaswamy and his team began an investigation in Google how the framework could be used to make the data common platform more accessible earlier this year.
Techcrunch event
San francisco
,
27-29 October, 2025
Google has also participated with a forest campaign, a non -profit organization that focuses on improving economic opportunities and public health in Africa, to launch a data agent. This AI tool uses MCP server on the tens of millions of financial and health data points in plain language.
A campaign contacted Google’s data common team with the prototype implementation of MCP on its own custom server. The conversation, Ramaswamy told Techcrunch, was the twist that inspired the team to create a dedicated MCP server in May.
However, the experience is not limited to a campaign. Data Commons MCP server’s open nature makes it compatible with any LLM, and Google has provided several ways to develop developers. A sample agent agent is available through the Development Kit (ADK) Collab notebookAnd the server can also be accessed through straight Gemini Cly Or using any MCP-compatible customer Pipi packageExample code is also provided on one Jethb repository,
