Microsoft says to be supporting Google’s newly released open protocol, which enables communication between AI “agents.”
Microsoft declared on Wednesday that Azure AI Foundry and Copilot Studio, two of its AI development platforms, would now support Google’s Agent2Agent (A2A) specification. In order to contribute to the protocol and tooling, Microsoft has also joined the A2A working group on GitHub.
The business stated in a blog post that it is “laying the foundation for the next generation of software — collaborative, observable, and adaptive by design” by embracing A2A and expanding on its open orchestration platform. “The best agents will work across models, domains, and ecosystems; they won’t reside in a single app or cloud.”
Google introduced A2A in early April, which enables agents—AI-powered, semi-autonomous programs—to collaborate across various services, apps, and clouds. Agents can invoke activities and share goals over the protocol. To ensure that agent collaboration takes place in a secure manner, developers are provided with a set of compatible components.
Agents developed using Azure AI Foundry and Copilot Studio will be able to use external agents for tasks, including agents hosted outside of Microsoft or made with other tools, once A2A support for the platforms is available. For instance, a Google agent might write the email invites while a Microsoft agent sets up a meeting.
According to the company’s blog post, “[C]ustomers can build complex, multi-agent workflows that span internal [agents], partner tools, and production infrastructure — while maintaining governance and service-level agreements.” “We’re supporting industry-wide efforts to implement shared agent protocols.”
Despite its flaws, businesses are increasingly investing in agentic technology as they seek to use it to increase productivity. A recent KPMG poll found that 65 percent of businesses are experimenting with AI agents. According to Markets & Markets, the AI agent market is expected to reach $52.62 billion by 2030, up from $7.84 billion in 2025.
Microsoft’s decision to enable A2A follows the release of Copilot Studio’s support for MCP, Anthropic’s standard for integrating AI with data storage systems. Earlier this year, several significant producers of AI models, such as Google and OpenAI, declared their intention to implement MCP.
