ARD Spec Lets AI Agents Discover Site Resources via a Single JSON File
Agentic Resource Discovery (ARD) is an open specification, released in 2026 under Apache 2.0, that allows websites to publish a machine-readable JSON manifest at /.well-known/ai-catalog.json listing all their agentic resources. Instead of AI agents inferring capabilities by crawling multiple pages, the manifest provides a structured index of MCP servers, APIs, A2A agents, and skill sets in one place. ARD functions purely as a discovery layer — it helps agents locate the right resource, which is then accessed through its own protocol separately. The specification builds on the AI Catalog data model maintained under the Linux Foundation and complements existing files like llms.txt, which maps content rather than agentic capabilities. As of a public census in June 2026, adoption remains limited, with even contributing companies yet to publish a discoverable ai-catalog.json, making early adoption a forward-looking move.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in