--- title: "Metabase AI" description: "Your AI assistant for working with data - from natural language questions, to SQL generation and editing, to AI generated summaries of charts." url: https://www.metabase.com/features/metabase-ai --- # AI exploration built on your permissions and semantic definitions ### AI-powered code generation and data exploration wherever you work. Every AI answer is inspectable, informed by your definitions, and governed by your permissions. [Try Metabase AI free](https://store.metabase.com/checkout?ai=1) [Deploy Metabase AI Open Source](https://www.metabase.com/start/oss) Trusted by 100,000+ companies ## Trustworthy AI analytics #### Inspect and verify every answer Every AI-generated answer links to the underlying query. See exactly which models and metrics were used, open it in the notebook editor or SQL, and edit if needed. No black boxes. #### Respects your existing permissions People only see data they're allowed to see. AI follows the same permission rules you've already set up. No new access layers. #### Informed by your semantic layer AI is aware of your metadata, definitions, and verified models that you define in [Data Studio](https://www.metabase.com/product/data-studio). Not generic SQL from a disconnected LLM. The richer your semantic layer, the better AI performs. #### Zero data movement Queries run against your warehouse. Metabot doesn’t ingest, store, or sync your data. Ideal for teams with strict boundaries or customer data segregation. #### Deploy and configure your way Plug in the AI provider you already use. Choose Claude, GPT, and other models from a [number of supported providers](https://www.metabase.com/docs/latest/ai/settings#choose-ai-provider). Switch anytime. Available for Cloud or self-hosted. ## AI access that stays under your control Decide who can do what with AI, and how much. #### Enable AI feature access per user group Control which groups can ask AI questions, use SQL generation, or access other AI features. Grant access granularly, so people get the tools that make sense for their role. #### Set usage limits by token or message count Keep AI spend predictable with usage limits per user-group and a total monthly cap. Choose whether limits reset daily, weekly, or monthly. #### Customize how AI appears in your organization Give Metabot a name and icon that fits your org. Set it up once and it's consistent everywhere your team uses it. #### System prompts to guide outputs Add instructions to shape tone, use the right naming conventions, or give context on your business — whatever helps it fit in better with your team and data. ## AI that isn't a separate meter Power Metabase AI with [your preferred model](https://www.metabase.com/docs/latest/ai/providers), including an open model you host yourself, and you pay that provider directly, with no markup from Metabase. Prefer not to manage a key? Metabase Cloud's managed AI service is a flat per-token rate with your first 1M tokens each month free. [See how AI pricing works](https://www.metabase.com/pricing) ## See who's doing what with AI, down to exact questions Track token spend, review conversation logs for security audits, and stay on top of any unexpected activity in the AI usage analytics dashboard. - #### Messages over time Per person or user-group. - #### Conversation logs per user What's asked, data accessed, and tool calls made. - #### Token usage Across people, user groups, and AI features. ## Explore your data with AI Ask questions, get summaries, and find what you need. #### Ask questions in plain language Chat with your data, get answers. Metabot builds queries using your semantic layer, returns charts and tables, and explains how it got there. How many new accounts were created last week? How many active users did we have in the last 6 months? Which plans are driving the most growth? What's our revenue this quarter? What's the average product rating? ##### New accounts created ##### Active users ##### Distribution of plans ##### Revenue per quarter ##### Average product rating #### One-click AI summaries One-click summaries of any chart or dashboard. Key takeaways without copying data into an external LLM. #### AI-powered semantic search Find dashboards, questions, and models by meaning, not just keywords. Search that leads people looking for 'revenue' to the right dashboard. Even if they call it 'earnings'. ## AI in your existing tools Get answers from Metabase in the tools your team uses every day. #### Official MCP server Connect to Metabase from your AI client. Query and interact with data in the same place you work with other integrated tools. Create dashboards, questions, and collections, get interactive charts as in-line results, run SQL, [and more](https://www.metabase.com/docs/latest/ai/mcp). #### Metabot in Slack @mention Metabot in channels or DMs. Ask questions, get charts, upload CSVs to Metabase, and set up subscriptions and alerts — all without leaving Slack. #### Agent Workspace API Programmatic access to build your own data agents on top of Metabase. Combine with the MCP server for end-to-end agentic workflows using your semantic layer. ## AI-assisted SQL and code Generate queries and debug faster. #### SQL generation Describe what you want in plain language. Get working SQL grounded in your data model, not hallucinated field names. #### AI-assisted debugging Troubleshoot failing SQL and get an explanation of what's wrong plus a suggested fix. ## Powered by your data model, metrics, and definitions AI only works as well as the context you give it. Your semantic layer defined in [Data Studio](https://www.metabase.com/product/data-studio) is the foundation that lets AI perform well. #### Define once, it's used everywhere Add business definitions to your Glossary. AI references them when generating answers, so 'ARR' means the same thing to Metabot as it does to your team. #### Surfaces existing work first Before running a new query, Metabot checks for dashboards or questions your team already built. Saves compute and reduces duplicated logic. ## Embed AI analytics in your product Give your customers natural language data exploration inside your app. The Metabase SDK includes AI components you can embed directly — scoped to each customer's data and permissions. [Explore Embedded Analytics](https://www.metabase.com/product/embedded-analytics) ### Frequently asked questions Can I choose my own AI provider? Yes. Bring your own key for any of the [supported AI providers](https://www.metabase.com/docs/latest/ai/providers), or host an open model yourself, and you pay that provider directly, with no markup from Metabase. Connect as many providers as you like, and switch anytime. If you don’t have a preferred AI provider, you can use the built-in model. Is AI available in open source Metabase? Yes. AI features are available in all Metabase editions, including the free open-source version. Some advanced capabilities like AI-powered semantic search and controls are available in Pro and Enterprise. Does Metabase charge for AI usage? Not when you bring your own key — you pay [your AI provider](https://www.metabase.com/docs/latest/ai/providers) directly and Metabase adds nothing on top. Metabot and the MCP server are included in every plan, including open source, so there’s no AI tier or per-seat AI fee. If you use Metabase Cloud’s managed AI service instead, it’s a flat per-token rate with your first 1M tokens each month free. How does Metabase AI handle my data? Queries run against your warehouse. Metabase doesn’t ingest, store, or sync your data. AI uses your semantic layer definitions to build queries. Your data stays where it is. Can I self-host Metabase with AI features? Yes. Metabase AI works in both cloud and self-hosted deployments. Self-hosted customers configure their own AI provider so everything stays in your environment. Can I embed AI features in my product? Yes. You can embed AI-powered querying components with Modular Embedding and the SDK. You can also use agent skills that teach LLMs about how to set up embedded analytics: for things like SSO implementation, version upgrades, and migrating from one type of embedding to another. Does Metabase have an MCP server? Yep! Connect your AI client to Metabase via the Metabase MCP server. Query, visualize, and explore Metabase data in the same place you work with other connected tools. Build dashboards, questions, and collections from your AI client, get interactive charts as in-line results, execute SQL, [and more](https://www.metabase.com/docs/latest/ai/mcp). Does Metabot work in Slack? Yes. Metabot in Slack lets your team ask questions, get charts, upload CSVs for analysis, and manage subscriptions and alerts without leaving their chat. Have more questions? We'd be happy to help. [Contact us](https://www.metabase.com/talk-to-a-person) ### AI that respects your rules. Start exploring. [Try Metabase AI free](https://store.metabase.com/checkout?ai=1) [Deploy Metabase AI Open Source](https://www.metabase.com/start/oss)