Catalog Capability¶

The Catalog capability enables agents to discover publishers and search content before accessing it. This fills the "discovery gap" — while Content Access assumes you already have a post ID, the Catalog enables scenarios like "find me AI newsletters under $10/month" that lead to subscription and access.

Following UCP's catalog pattern, UCAP splits the catalog into two independent capabilities that implementers can adopt separately:

Capability	ID	Description
Catalog Search	`dev.ucap.content.catalog.search`	Free-text search with filters and pagination
Catalog Lookup	`dev.ucap.content.catalog.lookup`	Direct lookup by publication or post ID

Motivation¶

Without a standardized catalog interface:

Agents cannot discover what content is available
Every integration requires custom discovery mechanisms
There's no way to search across multiple publishers

Data Structures¶

Publication¶

A publication represents a publisher's content collection (analogous to UCP's Product).

Field	Type	Description
`id`	string	Unique publication identifier
`slug`	string	URL-friendly identifier
`title`	string	Publication name
`description`	string	Publisher description
`url`	string	Canonical URL
`category`	string	Primary category
`price`	PriceRange	Price range (min/max across tiers)
`media[]`	array	Images, logos (first = featured)
`tiers[]`	array	Subscription options (first = featured)
`rating`	object	Aggregate agent ratings (optional)
`post_count`	integer	Total posts available
`language`	string	Primary content language (BCP 47)
`metadata`	object	Publisher-defined extensions

Post¶

A post represents a single piece of content (analogous to UCP's Variant).

Field	Type	Description
`id`	string	Post identifier (used as `post_id` in content access)
`external_id`	string	Source platform ID (e.g., Patreon post ID)
`title`	string	Post title
`excerpt`	string	Short preview blurb (1–2 sentences)
`teaser`	object	Publisher-authored content preview (see Teasers)
`published_at`	string	Publication timestamp (RFC 3339)
`min_tier`	string	Minimum tier required for access
`category`	string/array	Post category or tags
`media[]`	array	Featured images
`word_count`	integer	Content length
`reading_time_minutes`	integer	Estimated read time
`metadata`	object	Post-specific extensions

Context Object¶

The context object allows agents to specify preferences that affect search and display.

{
  "context": {
    "language": "en",
    "currency": "USD",
    "country": "US",
    "intent": "research on AI agent architectures"
  }
}

Field	Type	Description	Example
`language`	string	Preferred content language (BCP 47)	`"en"`, `"es"`, `"fr-CA"`
`currency`	string	Preferred price display currency (ISO 4217)	`"USD"`, `"EUR"`
`country`	string	Agent's operating region (ISO 3166-1 alpha-2)	`"US"`, `"GB"`
`intent`	string	Semantic description of search goal	`"find weekly AI newsletters"`

Publisher Teasers¶

Teasers are publisher-authored content previews that give agents enough context to evaluate relevance without requiring full access. Unlike the short excerpt (1–2 sentences), a teaser provides a substantial preview — typically the first few hundred words of the post, or a publisher-written summary optimized for agent consumption.

Teasers serve two purposes:

Better search results. Agents can use teaser content alongside context.intent to rank results more accurately.
Informed decisions. When an agent encounters paywalled content, the teaser helps it decide whether to recommend the content (and present a checkout URL) to the user.

Teaser Object¶

Field	Type	Description
`text`	string	Preview content (publisher-authored or auto-generated)
`word_count`	integer	Number of words in the teaser
`format`	string	Content format: `text`, `html`, `markdown` (default: `text`)

{
  "teaser": {
    "text": "The leaked GPT-5 benchmarks reveal significant jumps in reasoning, code generation, and multi-step planning. In this deep dive, we analyze what the numbers actually mean, compare them to Claude 4 and Gemini Ultra, and explore what this signals for the broader AI industry...",
    "word_count": 150,
    "format": "text"
  }
}

Teaser Configuration¶

Publishers control how teasers are generated for their content:

Setting	Type	Description
`teaser_mode`	string	`manual` (publisher writes each teaser) or `auto` (server generates from content)
`teaser_word_count`	integer	Default word count for auto-generated teasers (default: 150)

When teaser_mode is auto, the server extracts the first N words of the post content as the teaser. Publishers can override the auto-generated teaser on a per-post basis by providing a manually authored teaser.

Guidelines¶

Publishers SHOULD provide teasers for paywalled content to maximize agent discoverability
Publishers MAY use auto-generated teasers (first N words) or manually author them
Servers MUST include teasers in catalog search results and lookup responses when available
Servers SHOULD include teasers in Content Access 402 responses
Agents SHOULD use teaser content alongside context.intent to improve result ranking

Extensions¶

The catalog capability is designed for extension. Publishers and platforms can add domain-specific fields via the metadata object without protocol changes.

Planned Extensions¶

Extension	Capability ID	Description
Topics	`dev.ucap.content.catalog.topics`	Detailed topic/tag taxonomy
Freshness	`dev.ucap.content.catalog.freshness`	Content recency signals
Ratings	`dev.ucap.content.catalog.ratings`	Agent ratings and reviews
Recommendations	`dev.ucap.content.catalog.recommendations`	Personalized suggestions