---
title: "Agentic Resource Discovery: Let agents search"
description: "Agentic Resource Discovery (ARD) is a draft, open specification for discovering agent capabilities at runtime, addressing a model in which users and developers must pre-install, integrate, and maintain each MCP tool, skill, or A2A agent. It defines a static ai-catalog.json manifest and a dynamic POST /search registry API, using indexed signals such as publisher identity, representative queries, compliance attestations, and tags to move capability selection outside the LLM. Hugging Face's Discover Tool adapts Hub semantic search into ARD catalog entries for running Spaces, exposing skills, MCP servers, or raw Space metadata according to the requested media type and wrapping agents.md files into installable SKILL.md artifacts. The tool is available through the hf CLI, REST, and MCP, while planned work includes federation modes and Hub support for profile-level manifests."
---

# Agentic Resource Discovery: Let agents search

[Hugging Face](https://yomu.fyi/company/hugging-face) · ben burtenshaw, shaun smith · Jun 17, 2026

**Type:** Explainer

## Summary

Agentic Resource Discovery (ARD) is a draft, open specification for discovering agent capabilities at runtime, addressing a model in which users and developers must pre-install, integrate, and maintain each MCP tool, skill, or A2A agent. It defines a static ai-catalog.json manifest and a dynamic POST /search registry API, using indexed signals such as publisher identity, representative queries, compliance attestations, and tags to move capability selection outside the LLM. Hugging Face's Discover Tool adapts Hub semantic search into ARD catalog entries for running Spaces, exposing skills, MCP servers, or raw Space metadata according to the requested media type and wrapping agents.md files into installable SKILL.md artifacts. The tool is available through the hf CLI, REST, and MCP, while planned work includes federation modes and Hub support for profile-level manifests.

## Context

Agent capabilities currently follow an install-first, use-later model: developers hardcode MCP server URLs or users connect services through plugins. That approach does not scale to thousands of ad-hoc surfaces, while placing every tool description in an LLM context is constrained by context budgets and may not provide enough detail for reliable selection.

## Approach / What changed

ARD separates discovery from execution through a well-known static ai-catalog.json manifest and a REST registry API with POST /search. Hugging Face's Discover Tool adapts Hub semantic search into ARD entries, filters results to running Spaces, and returns skills, MCP server entries, or raw Space metadata based on the requested media type. It is accessible through the hf CLI, REST API, and an MCP endpoint.

## Takeaways

- ARD catalogs capabilities across federated registries using signals including publisher identity, representative queries, compliance attestations, and tags.
- Hugging Face Discover converts Space agents.md files into SKILL.md artifacts and can generate MCP catalog entries pointing to Gradio MCP endpoints for MCP-tagged Spaces.
- The specification uses a shared envelope for different artifact protocols and supports federation, allowing one registry search to surface capabilities hosted by another service.

**Tags:** [Developer Experience](https://yomu.fyi/topic/developer-experience), [Open Source](https://yomu.fyi/topic/open-source), [REST APIs](https://yomu.fyi/topic/rest-api), [Search](https://yomu.fyi/topic/search)

- Source: [Hugging Face](https://huggingface.co/blog/agentic-resource-discovery-launch)
- Source URL: https://huggingface.co/blog/agentic-resource-discovery-launch
- Ingested by Yomu: 2026-08-29T01:50:25.661Z

[Read original post](https://huggingface.co/blog/agentic-resource-discovery-launch)