Agentic Resource Discovery (ARD): What It Is and How It Works

Agentic Resource Discovery (ARD) is a draft open standard (version 0.9, May 2026) that enables domains to list their agentic resources in a standardized JSON document located at /.well-known/ard.json. Registries supporting ARD respond to POST /search requests to help discover these resources. ARD operates alongside the Model Context Protocol (MCP) and Agent2Agent (A2A) protocols, providing a complementary method for discovering AI agents and resources.

Key takeaways

• ARD is a draft open standard for listing agentic resources in a domain's /.well-known/ard.json file.
• Registries answer POST /search requests to enable discovery of ARD-listed resources.
• ARD complements MCP, which connects AI applications to tools via servers, and A2A, which enables agents to discover each other and exchange tasks.
• Wellknown indexes 1,466 records sourced from ARD catalogs.
• Wellknown publishes its own ARD manifest at https://wellknown.network/.well-known/ard.json.

What is Agentic Resource Discovery (ARD)?

According to the Wellknown glossary, ARD is a draft open standard (version 0.9, May 2026) under which a domain lists its agentic resources in a JSON document at /.well-known/ard.json. This document describes the resources available at that domain, enabling automated discovery by clients and registries.

Registries that support ARD respond to POST /search requests, allowing clients to query for agentic resources across multiple domains. This mechanism facilitates scalable and decentralized discovery of AI agents and tools.

How ARD fits alongside MCP and A2A

ARD is designed to complement two other open protocols:

• MCP (Model Context Protocol): MCP enables AI applications to connect to tools and data through servers exposing tools, resources, and prompts. MCP servers can be local packages or remote HTTP endpoints. MCP registries list servers with a server.json document.

• A2A (Agent2Agent protocol): A2A allows agents to discover each other through an Agent Card at /.well-known/agent-card.json and exchange tasks over JSON-RPC.

While MCP focuses on connecting AI applications to tools and A2A focuses on agent-to-agent communication, ARD provides a domain-level listing of agentic resources, enabling discovery at the domain scope. This makes ARD a complementary discovery layer that can be used alongside MCP and A2A.

ARD records in Wellknown's index

Wellknown indexes a total of 81,552 records, including AI agents, MCP servers, and tools. Among these, 1,466 records originate from ARD catalogs, as indicated by the ard_catalog source in the data. This shows that ARD is an active source of agentic resource metadata within the Wellknown ecosystem.

For comparison, other sources include:

• MCP registry: 49,210 records
• npm packages: 21,739 records
• PyPI packages: 20,136 records
• A2A cards: 2,839 records

This places ARD as a smaller but distinct source of agentic resource records.

Wellknown's own ARD manifest

Wellknown publishes its own ARD manifest at the URL https://wellknown.network/.well-known/ard.json. This manifest serves as a live listing of agentic resources indexed by Wellknown, formatted according to the ARD draft standard. It enables other systems and clients to discover Wellknown's aggregated agentic resources programmatically.

Summary

Agentic Resource Discovery (ARD) is a draft open standard that enables domains to publish a JSON manifest of their agentic resources at /.well-known/ard.json. Registries supporting ARD respond to POST /search requests to facilitate discovery. ARD complements MCP and A2A by providing domain-level resource listings alongside MCP's tool exposure and A2A's agent communication. Wellknown indexes 1,466 ARD-sourced records and publishes its own ARD manifest at https://wellknown.network/.well-known/ard.json.

This data is a snapshot as of 2026-11-02.