Google publishes Agentic Resource Discovery enterprise AI standard backed by Microsoft Salesforce Snowflake
Google $GOOGL has published an open standard, Agentic Resource Discovery, to let AI agents find and verify capabilities across business apps. Microsoft $MSFT, Salesforce $CRM, Snowflake $SNOW and Nvidia $NVDA back it. Anthropic and OpenAI are not among the initial supporters.
A group of major enterprise software and cloud companies has lined up behind a new technical standard meant to shape how AI agents work inside the workplace. On June 17, Google $GOOGL published the Agentic Resource Discovery specification, an open protocol for publishing, discovering and verifying AI capabilities across the web. The launch contributors include Microsoft $MSFT, Salesforce $CRM, Snowflake $SNOW, ServiceNow $NOW, Nvidia $NVDA, Cisco $CSCO, GitHub, GoDaddy $GDDY, Databricks and Hugging Face. The standard tries to answer three practical questions an AI agent faces at runtime: where the right capability lives, which one to use, and how to confirm it is safe to connect to. It defines a static manifest file that organizations host on their own domain plus a registry that indexes those catalogs and returns ranked matches to natural-language queries. The design is federated, so companies can run their own registries without a central authority. It is licensed under Apache 2.0 and builds on a Linux Foundation data model. The lineup is notable for who is missing. Anthropic and OpenAI, whose Claude and ChatGPT are pushing to become the single front door through which workers reach all their AI tools, are not among the initial supporters. The effort gives established software vendors a shared way to turn their large customer bases into AI distribution, using their existing apps as the entry point rather than ceding that ground to the newer AI labs.