Major AI Companies Announce Agentic Resource Discovery Specification (ARD)
Key Takeaways
- ▸ARD is an open discovery protocol that enables AI clients to dynamically find available external resources without pre-knowing what capabilities exist
- ▸The protocol operates exclusively as a discovery layer before invocation, complementing (not replacing) execution standards like MCP and A2A
- ▸Multiple independent discovery services can operate simultaneously with different trust models, ranking algorithms, and access policies tailored to their communities
Summary
A working group including Microsoft, Google, Hugging Face, GoDaddy, and other industry participants has announced the Agentic Resource Discovery Specification (ARD), an open protocol designed to help AI clients discover and match with external resources needed to complete tasks. ARD operates as a discovery layer that sits before resource invocation, enabling AI agents to query "What capabilities are available for this task?" and receive matching resources from discovery services.
Agentic resources encompass any external capability an AI can leverage—including agents, MCP servers, skills, APIs, plugins, workflows, canvases, and AI Catalog entries. Rather than acting as a replacement for existing standards like MCP, A2A, or AI Catalog services, ARD complements them by providing a unified discovery mechanism while allowing resources to be invoked through their native mechanisms.
The specification emphasizes a decentralized architecture where multiple independent discovery services can operate in parallel, each serving different communities and applying their own trust, ranking, and access policies. Organizations can run internal ARD discovery services for their own resources—similar to enterprise intranets—while public discovery services emerge across the web to serve different market segments, communities, and use cases based on their focus on quality, coverage, or specialization.
- Enterprises can run internal ARD discovery services for their own resources, enabling distributed resource management at scale
Editorial Opinion
ARD addresses a fundamental challenge in the emerging agentic AI ecosystem: resource discoverability at scale. By architecting discovery as a separate, composable layer, ARD enables flexibility and decentralization while allowing enterprises and communities to maintain control over trust and resource visibility. This collaborative open standard could become foundational infrastructure for the AI agent economy, much as open protocols like HTTP enabled the web.
