Eleven companies, including Google, Microsoft, GitHub, and Hugging Face, have published Agentic Resource Discovery (ARD). The open specification sets out how AI agents find and verify tools, skills, and other agents across the web.
The contributors released the draft spec on June 17, along with reference implementations from several of them. ARD is licensed under Apache 2.0 and builds on the AI Catalog data model maintained by a working group under the Linux Foundation. The full list of contributors also includes Cisco, Databricks, GoDaddy, NVIDIA, Salesforce, ServiceNow, and Snowflake.
What Does ARD Solve? #
The spec sets out to solve a coordination problem. Today, an agent has to be wired to each tool, MCP server, or API it uses ahead of time. As more companies publish their own capabilities, that pre-wiring stops scaling.
ARD moves discovery into a search step that happens at runtime. The shift mainly affects companies that publish tools and agents, not typical content sites for now.
How ARD Works #
ARD relies on two pieces, which the spec calls catalogs and registries.
An organization publishes a catalog, an ai-catalog.json
file hosted at a well-known path on its own domain, that lists the tools, MCP servers, agents, or APIs it makes available.
Registries then crawl those catalogs, index them, and answer discovery requests from agents in plain language.
Because each catalog sits on its own publisher’s domain, the spec uses domain ownership to verify who published it.
For production use, publishers can attach trust metadata so an agent or registry can confirm the publisher’s cryptographic identity before connecting. Once a capability is selected, ARD hands off and the agent connects directly using the tool’s own protocol.
Same-Day Implementations #
Several contributors shipped working tools built on the spec the same day.
GitHub introduced agent finder, which lets Copilot discover matching MCP servers, skills, tools, and agents from a chosen registry, with users controlling what gets connected.
Hugging Face released a Discover Tool that searches skills and MCP servers across ARD services. Cisco tied the spec to its AGNTCY Agent Directory, an open source project under the Linux Foundation.
The release continues a run of open specs aimed at the machine-readable layer of the web. Google published the Open Knowledge Format, a spec for sharing organizational knowledge between AI systems, two days earlier. The pattern across these efforts is the same. Each asks you to publish a structured file under your own domain so AI systems can use what you expose, without anyone wiring the connection by hand.
Where Google Fits #
Google’s role centers on Agent Registry, part of its Gemini Enterprise Agent Platform.
The company said Agent Registry will host and search agentic resources and handle enterprise governance. Native ARD support in the platform is planned for the coming months, which Google said would let organizations connect internal registries to the wider network. That support is not live yet, and ARD is a specification rather than a Google Search feature.
Why This Matters #
The split depends on what you publish. ARD is for publishers of callable capabilities, the APIs, MCP servers, and agents that software connects to. A company that publishes tools has a clear method for being found and trusted by agents. A typical content site has no clear action to take today.
The value of this effort is debated. Google’s John Mueller has argued that LLM systems can’t use files like llms.txt to distinguish one site from another, and advised focusing on current needs rather than future agent-oriented strategies. ARD targets tools and agents, not content, raising questions about building now for systems that may or may not generate traffic later.
Looking Ahead #
The spec is a v0.9 draft, and the contributors are inviting changes through the project’s GitHub repository. Its reach depends on registries that can crawl and index catalogs at scale, and that ecosystem is still in its early stages. Google’s Agent Registry support is months out.
If that network develops, the advantage mainly goes to companies offering tools and agents that others need. The early agentic-web features from Google hint at this. The immediate concern is whether your current platforms and tools will adopt ARD and what they will require you to publish.