CARTO is a location intelligence platform: it turns data about places, like stores, deliveries, customers or field visits, into maps and spatial analysis. It was founded in Madrid in 2012 by Javier de la Torre and Sergio Álvarez, and is now headquartered in New York, with offices in Madrid and Seville.
More than 2,200 customers and 350,000 users in over 50 countries use it, including Vodafone, BT, AXA, Unilever and JLL. Its most recent funding was a $61 million round in 2021, led by Insight Partners.
Built on your warehouse
CARTO’s approach is cloud-native: instead of copying data into its own system, it runs directly on the data warehouse where the data already lives. It connects to BigQuery, Snowflake, Databricks, Oracle, Redshift and PostgreSQL, or provides a warehouse of its own.
Hosted CARTO runs in several regions, Europe included, and it can also be self-hosted on any major cloud or on-premises.
What a BigQuery connection needs
CARTO offers three ways to connect BigQuery: signing in with a Google account, Workload Identity Federation set up by a Google Cloud admin, or a service account key, which CARTO suggests for testing rather than production. Every connection names a billing project, and the queries CARTO runs are billed to that project’s Google Cloud account, so CARTO recommends setting spending limits in BigQuery.
The permissions decide what works. Reading and listing data is the minimum, and is enough for read-only maps. Workflows also need to create a dataset for temporary tables, and imports, geocoding and saved results need to write and delete tables. Snowflake, Databricks and the other warehouses take the same levels, in each one’s own terms.
Getting sales and customer data into the warehouse
CARTO maps what’s already in the warehouse, so store sales or CRM accounts have to land there first. A one-off file goes in through Data Explorer, up to 5 GB, but an import from a URL happens once and doesn’t stay in sync with the source. Latitude and longitude columns are picked up automatically. Postcodes are often read as numbers unless the schema is set by hand, and rows whose coordinates can’t be read still import but stay off the map. Data that changes every day needs a scheduled feed into the warehouse, which a scheduled workflow can then build on.
Maps, workflows and agents
Builder makes interactive maps that technical and non-technical teams can explore and filter. Workflows lays out multi-step spatial analysis visually, runs it inside the warehouse, and can be scheduled hourly, daily, weekly or monthly. The Data Observatory adds thousands of ready-made datasets, from demographics to points of interest.
AI Agents, built inside a map or app, answer questions about the data in plain language. CARTO launched an MCP server in October 2025, and in May 2026 a set of tools for AI agents to work with it directly.
CARTO’s APIs
CARTO’s APIs cover maps, SQL queries, imports and exports, authenticated with API access tokens or OAuth. The location data services handle geocoding, including batches of up to 10,000 addresses, plus routing and travel-time areas through providers such as HERE, Google, Mapbox and TomTom.
Developers build custom map apps with CARTO’s JavaScript client and deck.gl, and a saved workflow can be run through the API, so another system can start it.
Basemaps on other sites
CARTO also publishes basemaps called Voyager, Positron and Dark Matter. They’re map backgrounds that websites and apps often draw their own data on, through Leaflet, OpenLayers or MapLibre. Since September 25, 2026, the raster tiles need an API key. Anyone can request one with an email address, with no CARTO account, and it goes on the end of the tile URL. Vector styles still load without a key for now, and maps made inside CARTO itself need no change.
Free use runs to 5 million tile requests a month for non-commercial sites and 1 million for commercial ones, with paid plans above that.
A key can be limited to named websites in the basemaps dashboard. Tiles requested from any other host are refused with a 403 and the map goes blank, and www.example.com and example.com count as different hosts, so a store locator on both needs both listed. On the free tier, the CARTO and OpenStreetMap credit has to stay visible on every map.
Where location comes from
CARTO maps what’s already in the warehouse, so the opportunities sit in what feeds it. Store sales, CRM accounts, delivery routes and field visits can land in BigQuery or Snowflake on a schedule, ready for a workflow that refreshes the map each morning, and a store locator on the website can run on CARTO’s basemaps.
Location analysis is only as good as the addresses and coordinates behind it, so the data needs to be ready before AI agents go on top of a map. A build-versus-hours estimate shows which feed is worth building first.