The SuggestAPI Knowledge Gateway now publishes an ARD-compatible catalog, strengthening SuggestAPI as The Discovery Layer for AI Agents and an Agent Search Layer for live commerce search, comparison, recommendation, and checkout handoff.
AI agents are becoming active participants in commerce. They search catalogs, compare products, evaluate policies, and help shoppers decide what to buy. To perform those tasks reliably, an agent needs more than access to an API. It needs a consistent way to find the right capability in the first place.
Google recently announced Agentic Resource Discovery (ARD), an open specification for publishing, discovering, and verifying AI capabilities across the web. ARD creates a shared discovery layer for tools, skills, APIs, MCP servers, A2A agents, and other callable services, regardless of the framework or provider behind them.
Today, we are announcing that SuggestAPI has implemented ARD.
Our Knowledge Gateway now publishes a live capability catalog at:
https://agent.suggestapi.com/.well-known/ai-catalog.jsonThis catalog gives ARD-compatible agents and registries a machine-readable description of the SuggestAPI capabilities they can discover and use.
Why agentic resource discovery matters
Today, many agent integrations depend on manually configured tools. A developer chooses an API, copies its description into an agent configuration, and decides in advance when the agent should use it. That model works for a small set of known tools, but it becomes difficult to manage as the number of agents and services grows.
ARD introduces a more open approach. A capability provider publishes an ai-catalog.json file at a well-known location on its domain. A client can fetch that catalog directly, or a registry can crawl and index it. The registry can then help agents find relevant capabilities from plain-language intent.
The distinction matters. ARD handles discovery, not execution. It helps an agent determine what capability exists and where its supporting artifact lives. The selected API, agent, or protocol still controls runtime authentication and interaction.
| Without a discovery standard | With ARD |
|---|---|
| Developers hardcode tool descriptions into each agent | Providers publish capabilities in a common catalog format |
| Agents are limited to tools configured in advance | Registries can index capabilities for dynamic discovery |
| Every platform invents its own discovery metadata | Entries use a consistent, artifact-neutral envelope |
| Publisher identity and trust information are handled inconsistently | Catalogs can include domain-anchored identity and trust metadata |
For agentic commerce, this provides a missing connection between an assistant’s intent and the live catalog capability that can fulfill it.
What SuggestAPI now publishes
SuggestAPI’s ARD catalog is hosted on agent.suggestapi.com, the domain of our Knowledge Gateway. It declares a domain-anchored host identity and describes each resource with an identifier, artifact type, capabilities, representative queries, version information, and supporting metadata.
The catalog currently advertises five resource surfaces:
| Resource | What an agent can discover |
|---|---|
| SuggestAPI Agent Registry | A registry interface for discovering commerce search, comparison, recommendation, and handoff capabilities |
| Tenant Search Surface | Tenant-scoped product and content search through a documented API surface |
| Agent Search Orchestration | Higher-level product retrieval, comparison, recommendation, and policy-aware answer generation |
| Agent Cart Handoff | A safe handoff plan containing merchant-controlled product, cart, and checkout resources |
| Guided Checkout Handoff | Guidance that routes a selected product toward the merchant’s existing checkout flow |
Each entry includes representative natural-language requests, such as finding a commerce search tool for an existing storefront or searching a merchant catalog for a specific type of product. ARD registries can use these examples to improve semantic matching between an agent’s request and the published capability.
How ARD fits the SuggestAPI architecture
SuggestAPI sits in front of an existing search stack. A merchant can keep Algolia, Typesense, Elasticsearch, Meilisearch, or a SuggestAPI index underneath while SuggestAPI adds query rewriting, normalization, routing, edge caching, analytics, and agent-ready responses. See Bring Your Own Search for the adapter model.
Our ARD implementation adds a standardized discovery layer to that architecture. This is what we mean by an Agent Search Layer: a shared interface that helps agents discover the right commerce search capability, connect to current catalog data, and receive structured results without coupling the agent to a specific search engine. The flow is straightforward:
- SuggestAPI publishes its capabilities. The Knowledge Gateway exposes its ARD catalog from the standard well-known path.
- An agent or registry discovers the catalog. It can fetch the file directly or find the published entries through an ARD registry.
- The agent selects the relevant surface. Descriptions, capability labels, and representative queries help match the user’s intent.
- The agent connects to the runtime API. SuggestAPI routes the request to the relevant tenant and underlying search source, then returns structured product or content results.
This allows discovery to remain open and interoperable while the merchant’s existing search infrastructure continues to serve as the source of truth.
Merchant control remains intact
Making a capability discoverable does not mean handing control of checkout to an agent.
SuggestAPI’s cart and checkout resources are explicitly designed as handoff surfaces. They can prepare product, cart, and checkout links, but they do not add products to a cart, place an order, or process a payment. Checkout remains platform-native, and payments stay on the merchant’s rails.
That separation is important. ARD helps an agent find the right capability. SuggestAPI helps the agent retrieve structured, current product information. The merchant’s storefront and commerce platform continue to control the transaction.
Discovery is the new SEO
A traditional storefront is designed to be discovered by people and indexed by search engines. An agent-ready storefront must also expose clear, structured capabilities that machines can find and understand.
SuggestAPI already helps shopping assistants search live product data instead of relying on brittle HTML scraping or stale model knowledge. ARD makes those capabilities easier to discover through an open standard. Learn more on the how it works page.
For merchants, the result is a practical path toward agentic commerce:
- Keep the search engine you already use. There is no requirement to replatform or rebuild the catalog.
- Return structured product data to agents. Assistants can search, compare, and recommend using current catalog information.
- Publish capabilities through an open discovery standard. ARD-compatible clients and registries can understand what the SuggestAPI Knowledge Gateway offers.
- Preserve the existing checkout experience. Product discovery can become agent-ready while the merchant retains control of cart, checkout, and payment.
The web became searchable because sites published information in formats that search engines could discover and interpret. The agentic web needs a similar discovery layer for tools and callable services. ARD is an important step in that direction. SuggestAPI is building The Discovery Layer for AI Agents, with an Agent Search Layer that connects agent intent to live, merchant-controlled catalog data.
Frequently Asked Questions About Agentic Resource Discovery
What is Google’s Agentic Resource Discovery specification?
Agentic Resource Discovery, or ARD, is an open specification for publishing, discovering, and verifying AI-facing capabilities across the web. It gives providers a standard way to describe tools, APIs, skills, agents, and registries so compatible clients can find them without relying on a manually configured list.
Although Google announced the project, ARD is an open, collaborative specification rather than a proprietary Google service. The current specification is published as a draft and is designed to work across providers, frameworks, and runtime protocols.
How does ARD work?
A provider publishes an ai-catalog.json file at /.well-known/ai-catalog.json on its domain. The catalog contains machine-readable entries describing the available resources, including their identifiers, artifact types, capabilities, descriptions, and supporting URLs.
An agent can fetch this catalog directly from a known provider. Alternatively, an ARD registry can crawl and index published catalogs, then help agents find an appropriate capability from a plain-language request. After selecting a resource, the agent connects through that resource’s native API or protocol.
Is ARD an agent runtime or communication protocol?
No. ARD is a discovery specification. It helps an agent answer questions such as what capability exists, who published it, and where its technical description can be found. It does not replace OpenAPI, MCP, A2A, or another runtime protocol. Authentication and execution remain the responsibility of the selected resource and its native protocol.
Does ARD make an AI resource safe to use?
ARD can improve trust and verification, but discovery alone is not a guarantee of safety. The specification allows publishers to include domain-anchored identifiers and optional trust metadata that registries or clients can verify. A client should still apply its own security policies, permission checks, authentication requirements, and risk controls before invoking a discovered capability.
What has SuggestAPI implemented?
SuggestAPI publishes an ARD-compatible catalog from its Knowledge Gateway at https://agent.suggestapi.com/.well-known/ai-catalog.json.
The catalog describes SuggestAPI’s agent registry, tenant search, agent search orchestration, cart handoff, and guided checkout handoff surfaces. Each entry includes structured discovery metadata such as a domain-anchored identifier, artifact type, capabilities, representative queries, version information, and supporting metadata.
What does ARD mean for SuggestAPI merchants?
ARD gives compatible agents and registries a standard way to discover the commerce capabilities published by the SuggestAPI Knowledge Gateway. A merchant can keep the search engine and catalog infrastructure they already use while SuggestAPI provides agent-ready product discovery on top.
The implementation does not require a merchant to replace Algolia, Typesense, Elasticsearch, Meilisearch, or another connected search source. SuggestAPI remains the layer in front, routing queries and returning normalized results.
Can an agent place orders or process payments through SuggestAPI’s ARD catalog?
No. Publishing a resource in an ARD catalog makes the resource discoverable; it does not grant an agent permission to complete a transaction.
SuggestAPI’s cart and guided checkout resources use a handoff model. They can return merchant-controlled product, cart, and checkout resources, but they do not add products to a cart, place an order, or process payment. The merchant’s existing commerce platform remains responsible for checkout and payment execution.
Does a merchant need to publish its own ai-catalog.json file?
Not necessarily. SuggestAPI publishes its platform capabilities from the Knowledge Gateway domain. A merchant’s integration can use those capabilities while keeping its existing storefront and search backend in place.
A merchant may choose to publish a separate ARD catalog on its own domain if it wants to advertise additional first-party agents, APIs, or tools. That decision depends on the merchant’s agent strategy and the resources it wants external registries to discover.
Is ARD finalized?
Not yet. The current ARD specification is version 0.9 and is labeled as a draft proposal. Implementers should expect the specification to evolve and should monitor the official project for schema changes, conformance guidance, and updated recommendations.
Where can developers learn more?
Developers can read Google’s ARD announcement, review the current ARD specification, and inspect SuggestAPI’s live ARD catalog. Teams evaluating the broader integration architecture can also review SuggestAPI’s Bring Your Own Search documentation.
Explore the implementation
You can inspect the live SuggestAPI ARD catalog, read Google’s ARD announcement, or review the current ARD specification.
If you want to make your product catalog easier for both shoppers and AI agents to search, join the SuggestAPI waitlist.
Your search stack stays. Your checkout stays. SuggestAPI becomes The Discovery Layer for AI Agents and the Agent Search Layer that makes your catalog discoverable to the agentic web.