On 17 August, Google's Agent2Agent protocol formally joined the Agentic AI Foundation, the Linux Foundation body that already governs Anthropic's Model Context Protocol. Two of the main interoperability standards in the open agent stack now sit under one neutral foundation, which counts more than 250 members, while each keeps its own maintainers and release schedule.
The two do different jobs. MCP standardises how an agent reaches a database, an API, or a file system. A2A standardises how one agent asks another to complete a task and return the result.
The adoption behind the governance news
Governance announcements are usually noise. This one follows adoption that is not.
MCP reached 97 million monthly SDK downloads by March 2026, comparable to the React package on npm but arrived at in sixteen months rather than three years. Every major provider now supports it natively: Anthropic, OpenAI, Google, Microsoft, AWS, Cursor, JetBrains. A2A reached version 1.0 in March 2026 with more than fifty technology partners including Atlassian, Salesforce, SAP, ServiceNow, and Workday.
When a standard reaches that level of support and then moves under neutral governance, the practical meaning is that it has stopped being a strategic asset for anyone. It is infrastructure.
What this removes as a place to compete
“If you describe your product mainly in terms of what it connects to, the standards have already made that description less valuable than it was when you wrote it.”
A meaningful number of products built over the past two years were, underneath, integration layers. The value proposition was connecting things that did not connect, and the moat was the effort involved in doing that.
Standards erode that moat directly. If reaching a system is a solved problem with a specification and an SDK, then having solved it is no longer worth anything, and the products whose entire value was connectivity are the ones that suffer.
If you describe your product mainly in terms of what it connects to, the standards have already made that description less valuable than it was when you wrote it.
This is the normal life cycle of infrastructure and it has happened before with every protocol that mattered. It is only notable here because of the speed. Sixteen months from introduction to React-scale adoption is not the pace at which these transitions usually run, and a lot of businesses were built during it.
Turn what you know into what you own.
Vibepreneur builds structured ventures from professional expertise, with positioning, launch assets, and growth systems included.
Join the WaitlistWhat it leaves
Everything above the connection. Once any agent can reach any system, the remaining questions are entirely about what should happen, which is a domain question rather than a technical one.
Which actions are permitted in this industry and which need a second pair of eyes. What the escalation path is when something is ambiguous. Which exceptions are routine and which mean stop. What a correct result looks like in enough detail that a machine could be measured against it.
None of that is in a specification and none of it generalises across industries. It is the accumulated operating knowledge of people who have run the process, which is the same asset that keeps turning out to be the durable one every time a layer below it gets commoditised.
The practical instruction
Do not build a proprietary connective layer, and do not treat integration breadth as a differentiator. Adopt the standards, spend as little time there as possible, and put your effort into the rules.
If you find yourself describing your product mainly in terms of what it connects to, the standards have already made that description less valuable than it was when you wrote it.
See what to build when everyone has the same tools for the wider version of this problem, and industry knowledge is the moat for what survives it. The system is organised around the layer that does not commoditise.