Working with tools

AI Agents transform natural language into spatial insights by invoking tools that can take geospatial action such as geocoding an address, executing a Workflows to identify best locations around geocoded location and add the output as new layer highlighting top locations.

Types of tools

Core tools

The «Core tools» are built-in tools by CARTO available immediately with no setup required. These handle common map interactions such as displaying a geocoded address on the map, zooming to a given location, filtering a layer or even adding new styled layers into the map.

MCP tools

You can provide your AI Agent with access to Workflows MCP Tools you have access to within your organization, expanding its capabilities. In CARTO you can create MCP Tools using Workflows, ensuring your AI Agent has access to valid and trustworthy tools that ensure your business logic is taken into account. Learn how to create Workflows MCP Tools in this section.

Once the required MCP Tools for your AI Agent are created, you can add them by clicking "Show tools" to expand the MCP Tool panel located on the right side. After that, you can review the Tool description, Input and Output details if present.

In the future CARTO will expand support to connect AI Agents to 3rd party MCP Servers. Please get in touch if you're interested in this upcoming feature.

Core tools overview

Core Tools are organized into four main categories. You don't need to know their parameters, just reference them by name in your Instructions, and the agent handles the details.

Map tools

These tools let the Agent control the map position as well as the map extracting area of interest.

Tool
Description

get_spatial_filter

Get current area of interest (viewport or polygon).

lds_geocode

Convert address to coordinates.

set_map_center_and_zoom

Zoom the map view to a specific layer extent.

set_map_center_and_zoom_to_layer

Zoom the map view to a specific layer extent.

get_map_coordinates

Retrieve the current map view coordinates and zoom level.

Layer tools

These tools let the Agent add, style and display layers on the map.

Tool
Purpose

add_layer

Visualize source data as new layer on the map.

style_layer

Style ai-generated layers.

set_layers_visibility

Remove ai-generated layers.

remove_layer

Removes a temporal ai-generated layer.

Widget tools

These tools let the Agent extract widget values and use them to filter data sources on the map.

Function Name
Description

filter_category_widget

Filter by categories (e.g., type, region)

filter_histogram_widget

Filter by distribution ranges

filter_range_widget

Filter by min/max values

get_formula_widget

Retrieve calculated metrics (KPIs).

get_category_widget

Get category breakdowns.

get_histogram_widget

Get distribution data.

get_time_series_widget

Get temporal patterns.

Extracting widget values using get_widget tools can retrieve up to 5,000 values.

Parameter tools

These tools let the Agent update parameter values on the map.

Tool
Purpose

set_sql_parameter_text

Update text parameter values.

set_sql_parameter_date

Update date parameter values.

set_sql_parameter_numeric

Update numeric parameter value.

set_sql_parameter_numeric_range

Update numeric range parameter values.

Data tools

These tools let the Agent generate SQL to add new sources on the map as well as executing queries against your map connection. Note this tools are off by default, you must enable Query sources capability in order to enable them. Learn how to configure capabilities in this section.

Tool
Purpose

add_source

Creates temporal ai-generated data sources from queries that can be used to render a new layer on the map.

remove_source

Remove a temporal ai-generated data sources.

execute_query

Run ai-generated query to provide insights.

Workflows Output tools

If a Workflows MCP Tool is provided to your Agent, your Agent comes with additional tools that let your Agent retrieve results from the Workflows execution. Learn about Workflows as MCP Tools in this section.

Tool
Purpose

add_source_from_workflows

Creates a temporal ai-generated layer using the fully-qualified table name output by a Workflows.

async_workflows_job_check_status

Check job status from Asynchronous Workflows execution.

async_workflows_job_get_results

Get the results from Asynchronous Workflows execution.

Last updated

Was this helpful?