What happened
The Agent2Agent Protocol (A2A) became a hosted project of the Linux Foundation’s Agentic AI Foundation (AAIF), moving from the foundation’s broader portfolio into the specialized governance structure that also houses the Model Context Protocol (MCP), AGENTS.md, goose, and agentgateway. A2A standardizes communication, discovery, delegation, and task exchange among independent AI agents across different vendors and frameworks.
Potential direction of power
Dispersion · Dispersion +1
The completed governance transfer strengthens the institutional infrastructure supporting a modular agent ecosystem. Placing A2A alongside MCP brings the principal standards for agent-to-agent communication and agent-to-tool connectivity under a common neutral governance home, potentially lowering integration and switching costs across providers and reducing the likelihood that interoperability develops primarily through proprietary platform interfaces.
The agent layer could evolve either as vertically integrated proprietary ecosystems or as interoperable infrastructure in which institutions can combine models, agents, tools, and platforms from multiple providers. Common governance of complementary open protocols improves the prospects for the latter and therefore preserves greater contestability at a potentially important emerging control layer.
Window of contestability
Opens
A2A was already under vendor-neutral Linux Foundation governance after Google donated it in June 2025; the August 2026 move does not newly remove Google control. Its significance is that A2A and complementary open agent-infrastructure standards are now governed within the same specialized foundation, whose membership includes competing AI, cloud, platform, and enterprise firms. Technical development remains openly governed, although proprietary platforms can still limit practical interoperability through identity, permissions, discovery, marketplaces, or extensions above the protocol layer.
