> For the complete documentation index, see [llms.txt](https://docs.carto.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.carto.com/carto-user-manual/settings/carto-ai.md). # CARTO AI CARTO AI adds AI capabilities across the CARTO platform. This includes [AI Agents](/carto-user-manual/ai-agents.md) in Builder for natural-language interaction with maps, along with other AI features where supported. ## Enabling CARTO AI CARTO AI is controlled from **Settings > CARTO AI** with two layers of toggles: 1. **Enable CARTO AI** (top-level toggle). This is the organization-wide master switch for all AI capabilities. When it is off, no AI feature is available to anyone in the organization, regardless of the granular settings below. 2. **AI Features** (per-feature toggles). Once CARTO AI is enabled, Admins can choose which specific AI features are available to users (see [AI Features](#ai-features)). To turn AI capabilities on for your organization, navigate to **Settings > CARTO AI** and toggle **Enable CARTO AI**.

Settings > CARTO AI, with the master toggle and the granular AI features

{% hint style="info" %} By enabling CARTO AI, you agree to the [CARTO AI terms and conditions.](https://drive.google.com/file/d/1QqidTLrDNqMbWg0D1RbHC3lkOSOHVXam/view?usp=drive_link) {% endhint %} ## AI Features Once CARTO AI is enabled, Admins can choose which specific AI features are available to users from the **AI Features and default models** section. Each feature has its own switch and its own default model selector, so you can enable only the capabilities you want available and pair each one with the model that best fits it. Currently, the following AI features can be toggled independently: * [**AI Agents in Builder maps**](/carto-user-manual/ai-agents.md): Editor users can create conversational agents that analyze and interact with maps, using available CARTO tools and MCP tools to provide insights directly in the interface. Enabled by default when CARTO AI is on. * [**AI Assistant in Data Observatory**](/carto-user-manual/data-observatory/accessing-and-browsing-the-spatial-data-catalog.md#ai-assistant-in-data-observatory): Users can find datasets in the Spatial Data Catalog by describing what they need in natural language, and the assistant applies the matching filters for them. Disabled by default. Admins need to enable it explicitly for it to appear in the Data Observatory. More features will become individually toggleable over time. ## Default model Each AI feature has its own default model, set from its model selector in the **AI Features and default models** section. This lets you pair every capability with the model that best fits it, so AI Agents and the AI Assistant in Data Observatory can each run on a different model. Available models in these selectors come from the CARTO-managed models or from any custom providers you have configured (see [Model Options](#model-options)). {% hint style="info" %} The default model is pre-selected when configuring or using AI-powered features. However, Editor users can override this selection and choose a different model as needed. {% endhint %} ## Model Options ### CARTO managed models (default) When you enable CARTO AI, these models are available immediately: * **claude-sonnet-4.6** *(default)*: Recommended default for geospatial use cases across CARTO AI. Strong reasoning, spatial analysis, and SQL generation with a good balance of quality, cost, and latency. * **claude-opus-4.8**: Most capable Claude model for the hardest geospatial reasoning and multi-step analytical tasks. Highest cost and latency. * **claude-opus-4.7**: Previous-generation flagship Opus, still available for organizations that have standardized on it. * **claude-opus-4.6**: Earlier-generation Opus, still available for continuity. * **gemini-3.1-pro**: Advanced Gemini model with strong geospatial reasoning and multi-modal capabilities. * **gemini-3.5-flash**: Fast and efficient Gemini model, ideal for simpler queries and high-volume interactions. ### Bring your own model Organizations can configure their own AI models from multiple providers, giving you full control over model selection, data residency, and cost management. #### Supported Providers
ProviderAuthenticationSupported Models
Google Vertex AIService Accountgemini-3.5-flash, gemini-3.1-pro, claude-opus-4.8, claude-opus-4.7, claude-opus-4.6, claude-sonnet-4.6, gemini-2.5-pro, gemini-2.5-flash, claude-opus-4.5, claude-sonnet-4.5
Google AI StudioAPI Keygemini-3.5-flash, gemini-3.1-pro, gemini-2.5-pro, gemini-2.5-flash
OpenAIAPI Keygpt-5.5-pro, gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.2-pro, gpt-5.2, gpt-5, gpt-5-mini, gpt-4o, gpt-4o-mini
AWS BedrockAWS Credentialsclaude-opus-4.8, claude-opus-4.7, claude-opus-4.6, claude-sonnet-4.6, claude-opus-4.5, claude-sonnet-4.5, claude-haiku-4.5
AnthropicAPI Keyclaude-opus-4.8, claude-opus-4.7, claude-opus-4.6, claude-sonnet-4.6, claude-opus-4.5, claude-sonnet-4.5, claude-haiku-4.5
Azure OpenAIAPI Key + Endpointgpt-5.5, gpt-5.4, gpt-5.4-pro, gpt-5.4-mini, gpt-5.2, gpt-5, gpt-5-mini, gpt-4o, gpt-4o-mini
Snowflake CortexPAT Tokenclaude-opus-4.8, claude-opus-4.7, claude-opus-4.6, claude-sonnet-4.6, claude-opus-4.5, claude-sonnet-4.5, claude-haiku-4.5, openai-gpt-5.4, openai-gpt-5.2, openai-gpt-5, openai-gpt-5-nano, openai-gpt-5-mini
Databricks Model ServingPAT Tokendatabricks-gpt-5.5-pro, databricks-gpt-5.5, databricks-claude-opus-4.8, databricks-claude-opus-4.7, databricks-claude-opus-4.6, databricks-claude-sonnet-4.6, databricks-claude-opus-4.5, databricks-claude-sonnet-4.5, databricks-claude-haiku-4.5, databricks-gpt-5.4, databricks-gpt-5.4-mini, databricks-gemini-3.5-flash, databricks-gemini-3.1-pro, databricks-gemini-2.5-pro, databricks-gemini-2.5-flash, databricks-gpt-5.2, databricks-gpt-5, databricks-gpt-5-nano, databricks-gpt-5-mini
Oracle Generative AIOCI Credentialsxai.grok-4.3, xai.grok-4.20-reasoning, xai.grok-4.20-non-reasoning, google.gemini-2.5-pro, google.gemini-2.5-flash
Custom (OpenAI Compatible)API KeyUser-defined (any OpenAI-compatible model)
#### OpenAI Connect directly to OpenAI's API. **Supported models**: `gpt-5.5-pro`, `gpt-5.5`, `gpt-5.4`, `gpt-5.4-mini`, `gpt-5.2-pro`, `gpt-5.2`, `gpt-5`, `gpt-5-mini`, `gpt-4o`, `gpt-4o-mini` **Configuration**: * **API Key** (required): Your OpenAI API Key * **Base URL** (optional): Custom API endpoint URL *** #### Anthropic Connect directly to Anthropic's API. **Supported models**: `claude-opus-4.8`, `claude-opus-4.7`, `claude-opus-4.6`, `claude-sonnet-4.6`, `claude-opus-4.5`, `claude-sonnet-4.5`, `claude-haiku-4.5` **Configuration:** * **API Key** (required): Your Anthropic API Key * **Base URL** (optional): Custom API endpoint URL *** #### Google AI Studio Connect to Google's AI Studio API. **Supported models**: `gemini-3.5-flash`, `gemini-3.1-pro`, `gemini-2.5-pro`, `gemini-2.5-flash` **Configuration:** * **API Key** (required): Your Google AI Studio API Key *** #### Google Vertex AI Connect to Gemini and Claude models via Google Cloud Platform. **Supported models**: `gemini-3.5-flash`, `gemini-3.1-pro`, `claude-opus-4.8`, `claude-opus-4.7`, `claude-opus-4.6`, `claude-sonnet-4.6`, `gemini-2.5-pro`, `gemini-2.5-flash`, `claude-opus-4.5`, `claude-sonnet-4.5` **Configuration:** * **Project ID** (required): Your GCP project ID * **Location** (required): GCP region (e.g., us-central-1) * **Service Account Credentials** (required): JSON credentials for a service account granted the **Vertex AI User** role (`roles/aiplatform.user`). See [Google's documentation](https://cloud.google.com/iam/docs/service-accounts-create) for creating a service account and granting roles. *** #### AWS Bedrock Access Claude models through AWS infrastructure. **Supported models**: `claude-opus-4.8`, `claude-opus-4.7`, `claude-opus-4.6`, `claude-sonnet-4.6`, `claude-opus-4.5`, `claude-sonnet-4.5`, `claude-haiku-4.5` **Configuration:** * **AWS Access Key ID** (required) * **AWS Secret Access Key** (required) * **AWS Region** (required): e.g., us-east-1 *** #### Azure OpenAI Access OpenAI models through Azure infrastructure. **Supported models**: `gpt-5.5`, `gpt-5.4`, `gpt-5.4-pro`, `gpt-5.4-mini`, `gpt-5.2`, `gpt-5`, `gpt-5-mini`, `gpt-4o`, `gpt-4o-mini` **Configuration**: * **API Base** (required): Your Azure OpenAI endpoint URL * **API Key** (required): Your Azure OpenAI API key * **API Key Version** (required): API version (e.g., 2025-01-01-preview) *** #### Snowflake Cortex Access AI models directly within your Snowflake environment **Supported models**: `claude-opus-4.8`, `claude-opus-4.7`, `claude-opus-4.6`, `claude-sonnet-4.6`, `claude-opus-4.5`, `claude-sonnet-4.5`, `claude-haiku-4.5`, `openai-gpt-5.4`, `openai-gpt-5.2`, `openai-gpt-5`, `openai-gpt-5-nano`, `openai-gpt-5-mini` **Configuration:** * **API Base** (required): Your Snowflake Cortex endpoint (e.g., `https://myorg-myaccount.snowflakecomputing.com/api/v2/cortex/inference:complete`) * **PAT** (required): Personal Access Token (PAT) *** #### Databricks Model Serving Access models through Databricks serving endpoints. **Supported models:** `databricks-gpt-5.5-pro`, `databricks-gpt-5.5`, `databricks-claude-opus-4.8`, `databricks-claude-opus-4.7`, `databricks-claude-opus-4.6`, `databricks-claude-sonnet-4.6`, `databricks-claude-opus-4.5`, `databricks-claude-sonnet-4.5`, `databricks-claude-haiku-4.5`, `databricks-gpt-5.4`, `databricks-gpt-5.4-mini`, `databricks-gemini-3.5-flash`, `databricks-gemini-3.1-pro`, `databricks-gemini-2.5-pro`, `databricks-gemini-2.5-flash`, `databricks-gpt-5.2`, `databricks-gpt-5`, `databricks-gpt-5-nano`, `databricks-gpt-5-mini` **Configuration**: * **API Base** (required): Your Databricks serving endpoint URL. * **PAT** (required): Databricks Personal Access Token *** #### Oracle Generative AI Access models through Oracle Cloud. **Supported models**: `xai.grok-4.3`, `xai.grok-4.20-reasoning`, `xai.grok-4.20-non-reasoning`, `google.gemini-2.5-pro`, `google.gemini-2.5-flash` **Configuration:** * **OCI User OCID** (required) * **OCI Tenancy OCID** (required) * **OCI Fingerprint** (required) * **OCI Private Key** (required) * **OCI Region** (required): e.g., us-ashburn-1 #### Custom (OpenAI compatible) Access models throgh your own LLM proxy or self-hosted endpoint that implementes the OpenAI-compatible chat completions API. **Supported models**: User defined, enter a comma-separated list of the model IDs your endpoint serves (e.g. model-1, model-2). **Configuration:** * **Base URL** (required): Your OpenAI-compatible endpoint URL (e.g., ) * **API Key** (required): Your own API key * **Models** (required): Comma-separated list of model IDs you serve {% hint style="info" %} When you **you bring your own provider, it replaces the CARTO-managed models**. Only the models from your configured providers will be available when creating AI Agents. {% endhint %} {% hint style="info" %} **Interested in other providers or models?** We're continuously expanding our AI provider support. If you'd like to see a specific provider or model added, pleas share your feedback with us. {% endhint %} ## Managing Models From the CARTO AI settings page, you can: * **Add providers:** Configure credentials for a supported provider to make its models available to the organization. * **Remove providers:** Disable a provider to remove its models from availability. Any models from that provider will no longer be selectable for new agents or AI features. * **Set default model**: Choose the default model for each AI feature. * **View available models:** See the full list of models currently available to the organization across all configured providers. ## Analytics Organization Admins can monitor how CARTO AI is being used across the organization from the **Analytics** tab in Settings > CARTO AI. See [CARTO AI Analytics](/carto-user-manual/settings/carto-ai/carto-ai-analytics.md) for details. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.carto.com/carto-user-manual/settings/carto-ai.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.