## Major tech companies back new protocol for AI agent resource discovery

A group of major technology companies, including Microsoft, Google, Cisco, Nvidia and Salesforce, is backing a new protocol intended to help AI agents discover the tools and services they need to do their jobs. The effort, called Agentic Resource Discovery, or ARD, is aimed at one of the emerging headaches of enterprise AI deployment: figuring out which systems an agent can use, where those systems live, and how access can be managed safely.

Enterprises that are experimenting with agentic AI often want those systems to take action across a wide range of internal resources. An agent might need to search engineering documentation, open a support ticket, review deployment history or query observability platforms. In many organizations, those capabilities are spread across separate registries and isolated business units. ARD is designed to provide a common layer that helps bring those resources together.

The protocol is built around two parts, catalogs and registries. Under the ARD model, organizations publish catalogs that describe the capabilities they make available. Registries then act as discovery services, scanning those catalogs so agents can find relevant tools and services. The goal is to create a standardized way for systems across a corporate domain to advertise what they offer and for agents to locate it without requiring custom integration for each environment.

Supporters of the initiative say the framework could reduce friction for companies trying to make AI agents more useful inside enterprise workflows. Instead of forcing developers to hard-code resource locations or build bespoke search methods for every internal system, ARD is intended to make discovery more consistent across different platforms and silos.

The specification is already available, and the project organizers are inviting organizations to publish their own catalogs using a quickstart guide on the ARD website. Companies can also join the project community on GitHub to contribute to the evolution of the specification.

The effort arrives as enterprise interest in agentic AI continues to grow, along with concern about governance and safety. Giving agents broader access to business systems raises obvious questions about control and oversight. By standardizing how resources are described and discovered, ARD attempts to address those concerns at the infrastructure level rather than leaving each organization to invent its own approach.

The project does not appear to replace existing enterprise systems or security controls. Instead, it is meant to sit above them as a discovery framework, helping agents identify available tools before they attempt to use them. In that sense, it is part of a broader push to make AI agents more interoperable inside corporate environments.

For now, ARD is still a specification in development, but it already has notable backing from some of the biggest names in technology. Whether it becomes a widely adopted standard will depend on how well organizations can use it to connect fragmented enterprise systems while still keeping access under control.