Connect your platform

Start here to bring CARTO into the AI platform you already use. Each guide covers how to set CARTO up in that platform — usually the CARTO MCP Server, and for coding agents also the Agent Skills and the CARTO CLI where they add value.

Before you start, get your MCP Server URL from CARTO Workspace under Developers > Workflow API & MCP Server, and decide how to authenticate — OAuth for the full tool surface, or an API Access Token for read-and-discovery work.

Chat and enterprise assistants

Connect over the MCP Server and work with CARTO in the conversation.

Claude

Claude.ai and Claude Desktop. Connects with just the URL.

ChatGPT

Custom connectors in Developer Mode.

Microsoft Copilot Studio

CARTO as a tool for your Copilot agents.

Google Gemini Enterprise

Custom MCP data store in AI Applications.

Snowflake Cortex

External MCP server for Cortex agents.

Databricks

MCP service through the Unity AI Gateway.

Oracle (OCI Generative AI)

Token-based MCP calling on the Responses API.

Coding agents

Connect over the MCP Server, and add the Agent Skills and CLI so the agent drives CARTO idiomatically and can script, automate, and administer.

Claude Code

MCP over the CLI, plus Skills and the CARTO CLI.

Codex

Remote MCP server in the Codex config, plus Skills.

GitHub Copilot

Remote MCP server in VS Code or the coding agent.

Antigravity

Remote MCP server in the agentic IDE.

What you can set up in each platform

Which platforms render maps inline?

CARTO's map tools render an interactive map directly in the chat only on platforms that support MCP Apps — Claude.ai, ChatGPT, and Claude Desktop today. On other platforms the agent still creates the map and returns its URL; open it in CARTO to view it.