Five enterprise vendors agree on how agents talk to software
Google, Microsoft, Salesforce, Snowflake and ServiceNow have all committed to Model Context Protocol as a shared standard for connecting AI agents to enterprise systems.
Five vendors who compete with each other on nearly everything have agreed on one thing: how an AI agent should connect to the software a business already runs. Model Context Protocol, originally Anthropic’s, is now a shared standard with Google, Microsoft, Salesforce, Snowflake and ServiceNow behind it.
Standards agreements are usually filed under integration and forgotten. This one has a second effect that is easier to miss. When every tool call an agent makes goes through one protocol, every tool call is describable in one vocabulary, which means it can be logged, replayed and audited in one place.
An agent that cannot explain its reasoning can still produce a complete record of what it did: which system it queried, with what arguments, and what came back. For a great many enterprise questions that record is the answer people actually needed. Not why the model believed something, but what it went and touched.