Summary
Google, Microsoft, Salesforce, Snowflake, and ServiceNow announced a formal alliance on July 13, 2026, to develop and promote a shared backend protocol standard for enterprise AI agent connectivity — a direct challenge to Anthropic’s Model Context Protocol (MCP), which has emerged as the de facto industry standard for how AI agents connect to enterprise tools, data sources, and each other.
The alliance aims to define open, vendor-neutral standards for agent-to-agent communication, enterprise data integration, and tool-use authorization under a neutral foundation’s governance. Anthropic’s MCP achieved rapid adoption in 2025-2026 largely by shipping early and becoming the default for developer tooling, giving Anthropic significant architectural influence over enterprise AI agent infrastructure. The coalition represents established enterprise platform vendors pushing back against that influence.
No formal specification has been released yet, but the alliance has committed to open governance. Enterprise organizations that have already built MCP-based integrations face potential fragmentation as competing standards emerge, and AI vendors will need to decide which protocol(s) to support natively in their products.
Source
Build Fast With AI — AI News Today: July 13, 2026
The Rundown AI — Enterprise AI Protocol Alliance Coverage
Commentary
This is the enterprise platform industry doing what it always does when a startup threatens to own a critical integration layer: form a committee. The difference here is that the committee is Google, Microsoft, Salesforce, Snowflake, and ServiceNow — collectively the dominant infrastructure of enterprise IT. Anthropic built MCP and got it adopted fast, but it’s a single company’s protocol, and the Fortune 500 is understandably uncomfortable depending on an AI safety lab for foundational integration standards.
The outcome will depend on execution speed and developer adoption, not committee size. MCP has momentum, tooling, and an existing ecosystem. The new alliance has market leverage but no spec. History suggests these situations resolve one of two ways: the alliance ships something genuinely better and the ecosystem migrates, or MCP becomes the entrenched standard and the alliance’s protocol becomes a niche alternative. Enterprises building production AI agent architectures right now should design for protocol abstraction — betting on a single winner before the spec exists is a recipe for painful refactoring later.