# platform_get_catalog Discover what data this platform serves: every category and dataset, with coverage, cadence, licensing posture, freshness, and the MCP tools that serve each dataset. | Field | Value | | --- | --- | | Category | Platform (`platform`) | | Exposure | public | | MCP tool | `platform_get_catalog` | | Documentation | https://www.agentdatasets.com/docs/tools/platform_get_catalog | ## Description Discover what data this platform serves: every category and dataset, with coverage, cadence, licensing posture, freshness, and the MCP tools that serve each dataset. When to use: call this first when unsure which category or tool answers a question, to inspect the human-readable universe and time-span description, or to check licensing posture. Sibling contrast: for data currency alone prefer `platform_get_freshness`; the catalog adds categories, coverage, posture, and tool routing on top of the same freshness snapshot. Catalog scope: it lists datasets and the tools that serve them. Identity/search helpers are described in the server instructions, not catalogued as datasets. The platform tools themselves are first-class registry rows under the launch-bar-exempt `platform` category: `platform_catalog` (served by this tool) and `platform_freshness` (served by `platform_get_freshness`), so the catalog describes its own metadata surfaces. What comes back: `data.categories` — the full fixed category list (key, display_name, status live|soon, description, exempt), always complete on every page; `data.datasets` — paginated dataset rows, each with `dataset`, `display_name`, `category`, `source_tag`, `source_name`, `license`, `posture` ("public"|"internal"), `redistributable`, `coverage`, `cadence`, `freshness` (status, stale, status_reason, last_refresh, latest_as_of), and `tools` (MCP tool name + one-line summary). `meta.source` is `platform_catalog`; `meta.as_of` is the freshness-check instant. Bounds + paging: default limit 50, max 100; all current datasets fit one default page; page by passing `pagination.next_cursor` as `cursor` until `has_more` is false. Errors: `bad_parameter` (malformed limit/cursor), `unavailable` (database unreachable). Example: `platform_get_catalog(limit=50)`. ## Parameters Input schema advertised to MCP clients and mirrored by the REST query string. | Parameter | In | Type | Required | Default | Constraints | Description | | --- | --- | --- | --- | --- | --- | --- | | limit | input | `integer` | no | 50 | min 1, max 100 | Catalog datasets to return per page (1-100); the default is 50, and the whole registry currently fits one page. | | cursor | input | `string \| null` | no | null | — | Opaque pagination cursor from a previous response's `pagination.next_cursor`; omit to start from the first dataset. | ## Provenance - `platform_catalog` — Curated internal registry (dataset catalog), Data-catalog metadata — categories, dataset descriptions, licensing posture, cadence, and ingest freshness; each listed dataset's own sources keep their originating-agency attribution (Catalog metadata is authored in-repo and carries no separate license; per-dataset licensing posture is stated on each row and governed by docs/licensing-register.md). See https://www.agentdatasets.com/docs/attribution.md Dataset registry: `platform_catalog`.