Choose theme
AI Resources
Agentic Resource Discovery
Agentic Resource Discovery, or ARD, is a draft specification for publishing, discovering, and searching agentic resources such as MCP servers, A2A agents, Skills, APIs, and workflows.
The current v0.91 proposal puts ARD entries in domain-hosted /.well-known/ard.json manifests or other advertised sources, then lets registries search them before a client uses the selected resource through its native mechanism. The older ai-catalog.json path remains an optional compatibility source. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.
What it is
A discovery layer for agent resources
ARD gives publishers a way to describe callable AI resources and gives registries a common search interface for returning matching capabilities before a client connects to the chosen tool or service.
Why it stands out
Discovery stays separate from invocation
ARD proposes a common way for registries to find and return agent resources while leaving credentials, permissions, safety checks, and execution to the systems that connect to them.
Availability
Draft spec plus implementation trail
Readers can inspect the ARD specification site, the public GitHub spec repository, Google and Hugging Face launch posts, and Hugging Face Discover as a working ARD client and server implementation.
Why it matters
What makes it useful
ARD addresses the step before invoking a tool: finding descriptions of available resources across registries. A publisher describes a capability; a client searches those descriptions before connecting to the selected service.
What to know
Where it fits
ARD fits publishers describing agent resources and registries helping clients find them. A client still uses the selected resource through its native protocol; credentials, permissions, and safety checks depend on the connected systems.
Notable points
What stands out
The v0.91 proposal uses /.well-known/ard.json and rel="ard" links for domain discovery, with POST /search for registry queries; ai-catalog.json remains optional compatibility. Hugging Face Discover implements search over Hub Spaces, Agent Skills, and MCP servers through REST and MCP paths.
Before using
What to review
The current document is a proposal; check schema and implementation compatibility before depending on its behavior.
Search results and publisher metadata do not supply credentials or permission to invoke a resource. Check the selected service and its native access requirements separately.
Check which registry is queried, who operates it, what it indexes, and how ranking, trust metadata, and access policies shape its results.
Check what publisher identity evidence is present and whether the client actually verifies it; metadata alone does not establish trust.
Review the selected resource's native access, privacy terms, rate limits, and logging after discovery hands off to invocation.
Reader fit
Who may find it relevant
Builders tracking how agents find tools, skills, and services across provider boundaries.
Teams comparing agent registries, internal catalogs, MCP servers, Skills, A2A agents, APIs, and trust metadata.
Builders who want a published specification and working reference implementation for testing discovery across agent formats.
Less relevant for readers looking for a finished consumer assistant, a model checkpoint, or a one-click decision about every returned tool.
Editorial note
Why LifeHubber lists it
For a publisher whose tools use domain-specific terms, v0.91’s JSON-LD @context and namespaces let those terms become discovery filter dimensions while retaining existing manifests. Define the terms before exposing them to registry search, so a client can ask for the relevant capability rather than only its name. Extension-term filtering depends on the registry: unsupported filters may be rejected, so check that the registry supports the terms you publish.
Source links
Source materials
Reader note
Before relying on this entry
LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.
What to explore next
Decide what happens after discovery returns a resource.
ARD helps a client find agent resources, then hands off to their native tools and protocols. These links cover reviewing a discovered skill and examining policy controls around later agent actions.
More in AI Agents
Keep browsing this category
Explore more AI agent projects.
Paperclip
paperclipai/paperclip
A self-hosted server and dashboard for coordinating agent teams through companies, goals, roles, issues, heartbeats, budgets, approvals, and persistent activity records.
AgentMemory
rohitg00/agentmemory
A self-hosted coding-agent memory server with session capture, keyword recall, optional local or remote embeddings, a viewer, and integrations through hooks, MCP, and REST.
Awesome GPT Image 2
freestylefly/awesome-gpt-image-2
A prompt-as-code library for GPT Image 2 with a categorized case gallery, reusable visual prompt templates, a live browsing site, and an installable agent skill for turning image ideas into structured prompts.
For project maintainers
Listed here? You can use the badge.
If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.
