AI Agents

CARTO AI Agents provide a powerful conversational interface that allows anyone, regardless of technical expertise, to ask questions in natural language and receive instant, actionable insights. This marks a fundamental shift beyond dashboards to a dynamic, intuitive way of exploring your geospatial data.

You can create agents directly in Builder and link them to your maps. This transforms static maps into interactive experiences where end-users can ask questions, explore data, and extract insights through conversation.

What is an AI Agent?

An AI Agent is a system powered by a large language model (LLM) that can interact with your data and tools to answer questions. Unlike a simple chatbot, agents can decide which tools to use, analyze results, and take multiple steps to solve a problem.

Agents have three components:

• Instructions: Define what the agent does, what it knows, and how it should behave. This is where you specify its purpose and expertise.

• Tools: The agent can access CARTO's built-in geospatial tools and your custom MCP tools for connecting to other systems.

• Model: The LLM that powers the agent's ability to understand questions and decide which tools to use.

---

Enable AI Agents in your organization

AI Agents are disabled by default. To enable them, navigate to Settings > CARTO AI and toggle Enable CARTO AI to enable them for your whole organization. Once enabled, any Editor user can create AI Agents in Builder maps.

To enable AI Agents in your organization you must be an Admin user.

---

Set up an AI Agent in Builder

Once AI Agents are available in your organization, you can start the creation of Agents directly in Builder. To start, create a new map or open an existing one. Then navigate to the AI Agents tab on the left pane and click Create agent. This will open the agent configuration dialog.

The Use Case field is required. Use it to explain what the map is for and what questions users will ask. For example, "This map shows retail locations. Users want to find stores near them and compare performance across regions." This helps the agent deliver relevant, accurate answers. Once you've filled in the Use Case, you're ready to test your agent. Everything else is optional.

You can also add custom instructions for more specific guidance. This is optional but recommended. Use it to add domain knowledge, define response style, or set boundaries on what the agent should and shouldn't do.

To enhance the user experience, you can set a welcome message that greets users when they open the agent, and add conversation starters—preset questions users can click to begin interacting with the agent.

Once you're ready, click Create Agent. This makes it available to Editors in your organization. To make it available for Viewers, toggle the setting in Map settings for viewers on the top banner. This will also make the agent available to anyone with the map link if the map is public.