Many of you are asking how we engineered the CARTO platform to run #Agentic #GIS without giving up on the enterprise governance, security & control requirements.
The short version, tier by tier:
🔹 Data: Nothing moves. Every query, every user click, and AI agent action on your data, compile to SQL that runs inside your own Google BigQuery, Snowflake, Databricks, Amazon Redshift, or Oracle. There's no additional database or client your data needs to sync with.
🔹 Models: You bring the ones your security team already approved. We support Anthropic, OpenAI, Google Vertex AI & Gemini, Snowflake Cortex, Databricks Model Serving, etc.. Your approvals, your keys, your contracts.
🔹 AI Agents and Assistants: Two different jobs. Your analysts create geospatial AI Agents running on your maps and custom apps, so any business user can ask spatial questions in plain language. The tools those agents call are your own authored analytical workflows, so every answer traces back to analysis a human can open and audit. AI Assistants sit inside our tools and make the analysts themselves faster at creating maps, assembling workflows, and wiring up agents.
🔹 Platform surface: CARTO for Agents (MCP Server, MCP Apps, CLI, Agent Skills) puts CARTO inside Claude, Gemini Enterprise, ChatGPT, Microsoft Copilot, GitHub Copilot, etc.
CARTO is the solution for teams looking for a modern GIS platform that can agentify their work, while running natively on their organization's data & analytics stack, and without sacrificing or duplicating any data governance and data access controls.