Use-case examples by platform

Each platform below has one concrete, end-to-end example you can adapt. They assume you've already connected your assistant to CARTO. Swap in your own connection, schema, and table names, and adjust the analysis to your data.

The examples are scoped to what each platform does best today. Assistants that render maps inline (Claude, ChatGPT) show visualizations in the chat; others return links you open in CARTO. For setup details on any platform, see Connect your platform.

Claude — build and refine a store dashboard

Claude connects with just your MCP Server URL and renders maps inline, so it's ideal for iterative map-making.

Use case: turn a raw stores table into a styled, shareable dashboard.

Explore my carto_dw connection and find my retail stores table. Create a Builder map showing the stores, colored by monthly revenue, with a legend. Then add a histogram widget on revenue and a category filter on region. Show me the map and give me the link.

Once you like it, keep going conversationally: "Make the top 10% of stores stand out," or "Set the map to public and give me a shareable link." Claude edits the existing map in place rather than rebuilding it.

ChatGPT — build and run an analytical workflow

ChatGPT (with a custom connector in Developer Mode) reaches the full CARTO surface, including authoring and running Workflows.

Use case: find which regions saw the most activity last month.

Build a CARTO Workflow that takes my events table, filters to the last 30 days, spatially joins it to my regions table, aggregates the event count per region, and saves the result as a new table. Then run it and summarize the top five regions. Finally, map the result colored by event count.

ChatGPT authors the DAG, validates it, runs it, reads back the output, and visualizes the result — a full analytical loop in one conversation.

Google Gemini Enterprise — enrich data for a market view

Gemini Enterprise connects as a custom MCP data store and reaches CARTO's full toolset.

Use case: understand the demographics around your locations.

Take my stores table and enrich each location with population and median income from the CARTO Data Observatory. Save the enriched table, then create a map showing stores sized by nearby population and colored by income. Give me the link to open it in CARTO.

Gemini finds the right Data Observatory variables, runs the enrichment, and produces a market-context map you can share with your team.

Microsoft Copilot Studio — explore data conversationally

Copilot Studio is quick for a maker to set up, and pairs naturally with business users asking questions of their data.

Use case: let a non-technical teammate explore the warehouse.

What connections do I have in CARTO? List the tables in my sales connection, then show me the row count and the columns for the transactions table. Which columns look like they hold location data?

Copilot returns clean, cited answers. From there, ask it to run a spatial query or open a saved map — a friendly entry point to CARTO for people who don't write SQL.

Databricks — query and prepare data with governance

Databricks exposes CARTO as a governed Unity Catalog service, so it fits teams that need auditing and access control.

Use case: prepare a spatial dataset for downstream analysis.

Using CARTO, run a spatial query that counts my sensor_readings points within each zones polygon over the last week, and return the busiest 20 zones. Then create a named source from that query so I can reuse it.

Every tool call is recorded in Databricks' usage tables, and you control which CARTO tools the agent can reach — useful when analysts share one governed connection.

Snowflake Cortex — query the warehouse through CARTO

Snowflake attaches CARTO as an external MCP server on a Cortex agent, so your CARTO tools live right next to your Snowflake data.

Use case: answer a spatial question and save a reusable source.

Through CARTO, query my Snowflake customers table and count customers within 5 km of each store in my stores table. Show me the stores with the most nearby customers, and save that query as a CARTO named source.

The Cortex agent runs the query through CARTO and creates a named source you can build on later — all inside Snowflake.

Oracle (OCI Generative AI) — read-only data discovery

Oracle's MCP calling is token-based, so CARTO exposes a read-and-discovery toolset there — perfect for exploration and reporting.

Use case: explore what's available before a deeper analysis.

What CARTO connections do I have? List the tables in my main connection, describe the schema of the parcels table, and run a query that summarizes parcels by land-use category.

Oracle's agent browses your connections, inspects schemas, and runs read-only queries — a solid foundation for building a larger analysis in a platform that supports the full toolset.

---

Adapt these to your data

Every example follows the same shape: describe your data, state the goal, and ask for the output (a map, a table, a summary, a reusable source). Agents work best when you name the connection and table, say which column to style or analyze by, and ask for a specific result. Start simple, then refine conversationally — the agent edits and reruns rather than starting over.

For the complete list of what agents can do, see the MCP Tools Reference.