
Microsoft will back Google’s open Agent2Agent (A2A) protocol with support for multiple products as the AI communication standard continues to gain steam.
The protocol, designed to allow cooperation between AI Agents on different platforms, will be supported on Microsoft’s Azure AI Foundry and Copilot Studio in public preview.
Now part of the A2A working group on GitHub, Microsoft said its support recognised “a new era of software design where intelligence is no longer tied to static interfaces or single applications.”
VP of Product for Azure AI Foundry Yina Arena and Copilot Studio CTO Bas Brekelmans said they were “laying the foundation for the next generation of software” as “the best agents won’t live in one app or cloud; they’ll operate in the flow of work, spanning models, domains, and ecosystems.”
See also: AI Agent protocol wars: IT’s A2A vs MCP vs ACP
The signs of Microsoft’s A2A support could be seen a few weeks ago, when it revealed Azure AI Foundry’s agent development kit, Semantic Kernel, would ‘speak A2A’ and shared an initial contribution to the protocol’s repository highlighting its then lack of a “packaged A2A library”.
In its official announcement, Microsoft said a new sample in Semantic Kernel could demonstrate how local agents can collaborate using A2A ahead of its public preview in Foundry and Copilot Studio “soon”.
A protocol boom
Google’s standard is still in development but a splashy announcement last month touted the backing of Salesforce, MongoDB, KPMG and McKinsey, though Microsoft’s support is likely the best sign for its industry-wide adoption so far.
In fact, the latter said its contributions would “accelerate development and adoption” of A2A across the industry and described it as an “important step” in realizing its “vision for the agentic future.”
See also: Microsoft conduct "harming competition in cloud services"
Microsoft’s support also comes just a few months after it introduced the Anthropic-developed Model Context Protocol (MCP) to Copilot Studio, facilitating connections between its AI Agents and third-party data sources.
With both A2A and MCP, the company has been cautious about appearing stuck with its protocol choices though, telling customers it would “support the protocols, models, and frameworks” that mattered to them.
Many companies have shared similar sentiments as they test the agentic protocol waters, Veeam’s CTO told The Stack during its recent VeeamON event that, despite revealing support for MCP, “we are not necessarily exclusive to that.”
Like Microsoft, Amazon, an investor in Anthropic, has been supporting MCP for a few months across AWS’ Amazon Bedrock and Amazon Q CLI products.
So far though, it remains unclear if it will also decide A2A is the way to go, with other protocols including the CISCO-backed ACP offering similar services.
Some AWS customers have asked for a roadmap on the issue, but the company’s re:Post AI Agent told at least one person “there are no specific announcements or roadmaps” related to AWS and agentic protocols, not yet anyway.
Sign up for The Stack
Interviews, insight, intelligence, and exclusive events for digital leaders.
No spam. Unsubscribe anytime.