September 2026 Summaries

Public sector GIS teams facing Esri product retirements and constrained staffing can modernize by auditing applications before migrating them, running spatial analysis directly in centralized cloud warehouses, consolidating permissions and audit trails, enabling self-service for routine questions through AI Agents, and integrating spatial data into existing BI and AI tools. The approach emphasizes identifying unused or outdated apps, eliminating duplicated datasets and synchronization delays, and using existing security controls such as role-based access, row-level security, and SSO to improve governance. GIS specialists retain responsibility for authoritative data and high-stakes outputs, while reusable workflows can allow nontechnical staff to answer common spatial questions more quickly. CARTO presents its cloud-native platform as a way to support these practices within warehouses including BigQuery, Snowflake, Databricks, Redshift, Oracle, and other organizational systems, without copying underlying data.

CARTO is participating in an 18-month European Commission DG DEFIS study, launched in Brussels in September 2026, to assess how core geospatial data is governed, produced, licensed, and shared across all 27 EU Member States and to develop options for a common policy framework. Led by Evenflow with ConsultingWhere, CARTO, and BHO Legal, and supported by organizations including the Open Geospatial Consortium and EuroGeographics, the project will consult data providers and users while examining how authoritative national, local, commercial, and EU space data can be integrated. CARTO leads the technology-focused option and will build an open-source, platform-independent demonstrator showing potential uses of integrated national geospatial and EU space data. Its proposed approach emphasizes AI-ready data with clear metadata and traceability, data sovereignty through federated stewardship, scalable cloud-native access, and interoperability through open standards, compatible formats, shared definitions, and transparent access rights.

CARTO has named Ubilabs its exclusive partner and reseller for Germany, Austria, and Switzerland, combining CARTO’s cloud-native Agentic GIS platform with Ubilabs’ 20 years of geodata implementation experience. The partnership aims to help DACH organizations turn existing location data into faster, more accessible decision-making by enabling AI agents, geospatial workflows, and interactive maps to operate directly within customers’ cloud data warehouses under existing governance. Ubilabs will assist customers from identifying use cases and establishing technical foundations through production deployment, supporting applications such as site selection, logistics planning, risk assessment, and self-service spatial analysis.

Esri is retiring several ArcGIS products between 2026 and 2027, including Classic Story Maps and Map Viewer Classic in Q1 2026, Configurable Apps in ArcGIS Online from February 2026, ArcGIS Workforce online in Q4 2026, and Web AppBuilder applications by Q2 2027. Because Web AppBuilder and Configurable Apps require manual rebuilding rather than automatic conversion, organizations must inventory apps, assess active use, rebuild and test needed tools, update links, and retrain users, creating potentially substantial staff costs alongside changes to Concurrent Use licensing. The piece recommends using the retirements to eliminate unused apps and decide whether remaining tools should be replicated on Esri’s replacement products or redesigned around broader organizational data, access, and workflow needs. It presents CARTO as an alternative cloud-native, warehouse-based GIS platform, claiming its migration tools can help move ArcGIS content into data warehouses and recreate maps, workflows, and some applications while retaining user approval over changes.

CARTO’s MCP Server can be connected to Databricks Genie Code to give AI agents access to CARTO’s map-building, geospatial analysis, workflow-authoring, and administrative capabilities while queries continue to run in the customer’s Databricks environment against Unity Catalog tables. The integration uses authenticated user credentials and existing permissions, row-level controls, and activity logging, with no data copied to external geospatial databases. CARTO Agent Skills complement the MCP tools by guiding agents in selecting and applying appropriate spatial methods. Setup requires a compatible Unity Catalog workspace, CARTO OAuth or API credentials, registration through Databricks AI Gateway, tool loading, and activation of both the connector and skills in Genie Code. Example uses include geocoding customer addresses, creating interactive retail-performance maps, performing routing and isoline analyses, and converting analyses into reusable CARTO Workflows that compile into native Databricks workflows and can be scheduled, parameterized, shared, or exposed to other agents.

CARTO’s MCP Server can be connected to Snowflake CoWork to let AI agents create and revise maps, run geospatial analyses, build reusable CARTO Workflows, and perform selected administration tasks directly on governed Snowflake data without exporting copies. Queries execute in each user’s Snowflake context after individual CARTO consent, preserving existing roles, row-level access policies, masking, and auditability. Setup requires Snowflake ACCOUNTADMIN access, a confidential CARTO SPA OAuth client, connector configuration in Snowflake, and agent permissions and skills. Examples include mapping retail performance, publishing parameterized drive-time workflows as reusable tools, and generating drive-time coverage maps through CARTO’s Location Data Services and Snowflake Analytics Toolbox. The related CARTO CLI and MCP integration for Snowflake CoCo support more scripted, bulk, administrative, and application-development work, enabling users to build spatial apps and versioned analytical workflows on Snowflake-managed data.

CARTO’s guide explains how to connect its Model Context Protocol (MCP) Server to Microsoft Copilot Studio, enabling Copilot agents in Microsoft 365 or Teams to create and edit maps, build and run CARTO Workflows, query spatial data, perform geocoding and routing, locate existing assets, and support platform administration through conversational prompts. Actions run under the authenticated CARTO user’s existing permissions, including row-level security, execute in the organization’s connected data warehouse, and are logged in CARTO Activity Data. Setup involves creating a CARTO SPA OAuth client, manually configuring the MCP tool with OAuth 2.0 in Copilot Studio, adding the generated callback URL, signing in, and publishing the agent; a read-focused API-token option is also available without map or workflow authoring. The guide recommends enabling generative orchestration, disabling dynamic discovery, and optionally providing CARTO Agent Skills to improve geospatial decision-making. Examples show Copilot producing a Chicago transit-access map with cartographic styling and creating an auditable, reusable workflow pipeline, while the integration is presented as a way to broaden access to governed geospatial analysis and reduce reliance on specialist GIS teams.

At the Open Visualization Collaborator Summit at ETH Zürich, CARTO founder Javier de la Torre presented an evaluation of how seven advanced AI models recommend, generate, and interact with deck.gl maps, finding that while most can produce functional and often polished visualizations, their code commonly relies on outdated deck.gl versions, deprecated integration patterns, and stale official examples. The tests showed broad agreement that deck.gl is suited to large-scale data visualization, MapLibre to basemaps and React applications, and Leaflet to simple maps, but models frequently lacked awareness of deck.gl 9.4 and struggled with confusing MapLibre integration options, skewed-data styling, legends, and silent failures in declarative JSON specifications. CARTO proposed making a schema-backed @deck.gl/json v2 the primary agent interface, with validation, conversion reports, state read-back, patch editing, data-source support, and standardized color expressions, while also improving documentation through llms.txt files, Markdown pages, agent skills, and updated examples. The presentation additionally called for vis.gl to formalize AI-assisted contribution practices under the OpenJS policy and determine how automated development, review, verification, and security should operate as agents become both major users and contributors to deck.gl.

CARTO has released guidance for connecting its MCP Server to Claude as a custom connector, enabling users to create and revise maps, run geospatial analyses, build CARTO Workflows, geocode, route, access connected data, and perform platform administration within Claude conversations. The integration uses authenticated CARTO accounts, executes queries in customers’ connected data warehouses, preserves existing IAM and row-level security controls, and records actions in CARTO Activity Data. Setup requires adding the CARTO MCP Server URL in Claude’s connector settings and signing in, without registering an OAuth client. Examples demonstrate using Claude Cowork to create a reusable Denver telecom coverage workflow that identifies buildings beyond LTE tower service areas and to produce a decision-oriented map of coverage gaps. For repository-based application development and more complex analysis, CARTO also supports Claude Code through its CLI and Agent Skills, allowing users to build spatial applications and versioned analytical workflows.

CARTO’s guide explains how to connect its MCP Server to Google Gemini Enterprise so employees can use conversational AI to create maps, run geospatial analyses, build reusable CARTO Workflows, and perform selected platform administration tasks. Requests are authenticated as the individual user, run directly in the organization’s connected cloud data warehouse, retain existing IAM permissions and row-level security, and are logged through CARTO Activity Data. Setup requires creating a CARTO SPA OAuth client, adding the CARTO MCP Server as a custom MCP data store in Google Cloud AI Applications, enabling selected tools, and linking it to a Gemini Enterprise application. Examples include a San Francisco EV-charging accessibility dashboard and a Denver cell-tower planning workflow, illustrating how agents can produce visualizations and auditable spatial pipelines from warehouse data. The guide also notes that Gemini Enterprise currently cannot display maps inline, so agents should return shareable CARTO URLs, while unused MCP data stores should be removed because they are billable.

Portolan is presented as an open, cloud-native approach to spatial data infrastructure designed to improve how authoritative geospatial data is published, discovered, queried, and governed. It responds to limitations of traditional portals and API-based services by enabling publishers to store data in open formats such as GeoParquet, Cloud-Optimized GeoTIFF, PMTiles, and potentially Apache Iceberg, allowing users and applications to query files directly from publisher-controlled object storage. The approach combines standardized STAC catalogs, human-readable documentation, agent-oriented guidance, validation tools, and a registry of independently hosted catalogs to support federation without centralizing datasets. The text argues that AI agents can make these distributed data sources more accessible by discovering relevant datasets, performing reproducible spatial analyses, and returning sourced results, while open formats and separable choices of storage, compute, AI models, and semantics support data sovereignty. CARTO SDI is described as a commercial implementation built around the open Portolan catalog, adding catalog interfaces, natural-language AI access, governance, permissions, telemetry, quality controls, monetization, and support for legacy GIS services while leaving data in the publisher’s chosen storage environment. The initiative remains early, with approximately 20 registered catalogs, and seeks broader adoption, format support, client integrations, and independent implementations.

CARTO’s PlacePulse Embeddings, developed with Applied Geographic Solutions, represent locations across the United States as 256-dimensional vectors that combine demographic, commercial, and environmental characteristics into searchable digital fingerprints. The approach is presented as a way for retailers and other location-based organizations to identify what their highest-performing sites have in common without manually selecting and weighting hundreds of individual variables. By using cosine-similarity searches on embeddings for top-performing store catchment areas, organizations can create a “retail fingerprint,” map areas with similar profiles, rank expansion opportunities from Prime to Low Potential, and incorporate constraints such as distance from existing stores to assess infill or new-market opportunities. Available through CARTO’s Data Observatory and Workflows, the reusable embeddings can support broader applications including site selection, market segmentation, prediction, and identifying promising locations for businesses such as banks, clinics, dealerships, and gyms.