carto-find-spatial-data

carto-find-spatial-data

The CARTO Data Observatory (DO) is a catalog of curated external spatial datasets — demographics, points of interest, environmental data, mobility, financial — that you can subscribe to and have materialized into your own data warehouse. Instead of the user hunting down a US census shapefile, this skill answers: "what's already curated in the DO that would help here?"

This skill covers discovery and subscription. Once subscribed, the data lives in the user's warehouse and is queryable like any other table — see `carto-query-datawarehouse`.

Access-path routing. With the CARTO MCP server attached, route discovery through search_data_observatory (method: list_datasets|list_filters|search_variables|get_dataset|sample) — available even on token-authenticated sessions — and subscription management through manage_data_observatory_subscriptions (method: list|subscribe|unsubscribe), which needs an OAuth session. Fall back to the carto do CLI when the server isn't attached, for scripted use, or to subscribe from a token session. The evaluation guidance below (coverage, licensing, cadence) applies on every path. Detection signals: `carto-basics/references/access-paths.md`.

When to use this skill

• The user asks for "demographics", "points of interest", "boundaries", "mobility data", "weather", or any other category of external spatial data.
• The user has internal data and wants to enrich it with external context (e.g. spatial-join their stores against census tracts).
• The user is exploring what's available before committing to a paid subscription.

If the user already has the dataset in their warehouse, jump to `carto-explore-datawarehouse` instead — DO is for sourcing new external data.

Quick reference

MCP (search available on any session; subscribe needs OAuth):

search_data_observatory { method: "list_datasets", keyword: "census", country: "usa" }
search_data_observatory { method: "get_dataset", dataset_id }        # inspect before subscribing
manage_data_observatory_subscriptions { method: "list" }
manage_data_observatory_subscriptions { method: "subscribe", dataset_id, connection, destination }

CLI fallback (no server / scripted / token-session subscribe):

carto do search --keyword "census" --country usa --json
carto do get <dataset-id> --json
carto do subscriptions list --json
carto do subscribe <dataset-id> --connection carto_dw --destination my_project.do.census_tracts

What's in this skill

Always-on guidance

• Free vs paid: many DO datasets are free (US Census, OpenStreetMap-derived, NaturalEarth). Premium datasets (Spatial.ai, Mastercard, AirSage, etc.) require a CARTO subscription contract — get_dataset / carto do get shows the licensing and pricing tier.
• Subscription lands data into the user's warehouse. It's not a remote view — actual rows are written to a destination table the user pays warehouse storage for. Plan destination naming accordingly.
• Spatial extension required. DO datasets often include H3 cells / geometry types that depend on CARTO's spatial extension being installed in the destination warehouse. If subscribing fails with "function not found", check the extension is installed — see `carto-manage-platform`.
• Don't recommend a DO dataset without inspecting it first (get_dataset / carto do get). Its metadata reveals coverage (US-only, EU-only, global), update cadence (monthly, yearly, real-time), and licensing — all critical to whether it fits the user's question.
• DO is external; named sources are internal.* Named sources (covered in `carto-explore-datawarehouse/references/named-sources.md`) are the user's saved queries; DO is the curated catalog of third-party data. Don't confuse the two.