---
title: "Self-service BI and AI analytics for your team | Metabase"
description: "Metabase is open source business intelligence. Your team explores data, asks Metabot or asks Claude and ChatGPT through the MCP server, all on the metrics and permissions your data team set once."
url: https://www.metabase.com/product/business-intelligence
---

Business intelligence

# Your team answers its own questions, on numbers everyone agrees on

People click into a chart, ask Metabase's AI, or ask the AI they already use, like Claude or ChatGPT. Every answer comes from the metrics and permissions your data team set once.

[Get started](https://store.metabase.com/checkout) [Install open source](https://www.metabase.com/start/oss)

Rolling it out across a larger company? [Talk to sales](https://www.metabase.com/sales)

## Everyone asks their own way, and gets the same answer

Your data team stops being the queue for every number. People ask in plain English, click through charts or write SQL, and get the answer themselves.

- **Ask in plain English.** [Metabot](https://www.metabase.com/features/metabase-ai) answers in Metabase or in Slack, and shows the query behind every answer.
- **Ask from their own AI.** Through the [Metabase MCP server](https://www.metabase.com/features/mcp-server), Claude, ChatGPT or Cursor signs in as the person asking and answers with Metabase charts.
- **Point and click.** Build a question in the [query builder](https://www.metabase.com/features/query-builder) without writing SQL, then click any chart to [drill through](https://www.metabase.com/features/drill-through) to the next question.
- **SQL when they want it.** Analysts write SQL in the [SQL editor](https://www.metabase.com/features/sql-editor), and Metabot can draft or fix it.

[Explore AI governance](https://www.metabase.com/product/ai-governance)

Chat with Metabot

Why are signups down in 2026?

**Signups are down across every source in 2026,** so it’s broad, not one channel. Organic still leads with 62, then Google with 30.

Data Sources

Signups Published metric · Library

Every source is below its 2025 pace, so even the top channel is down.

Want me to compare @New Accounts by Source, 2026 vs 2025?

Ask about your data

## Define a number once, and every dashboard, Metabot answer and agent uses it

Your data team defines tables, metrics and business terms in [Data Studio](https://www.metabase.com/product/data-studio), your semantic layer. Dashboards are built on those definitions, and AI looks for them before it touches raw tables. Change a definition once and everything built on it follows, and Metabase [lists what the change would break](https://www.metabase.com/features/data-lineage) before you save, so nobody opens a broken chart.

## When a team needs its own tool, your agent builds it on the same numbers

A renewal desk, an approval queue, a CRM for one team: ask your coding agent and it builds a [data app](https://www.metabase.com/product/data-apps) (beta) that lives in Metabase, on your semantic layer, sign-in and permissions. Nothing new to host or secure.

- ### The same numbers as your dashboards

  Apps read your published tables and metrics, so the app and the dashboard can't disagree.

- ### Each person sees their own slice

  People sign in with Metabase and see only what their groups allow, in the same app.

- ### Changes go through your actions

  Apps write only through actions admins control, and every app shows up in the dependency graph before you change a metric it uses.

[Explore data apps](https://www.metabase.com/product/data-apps)

## One place for the reports people use every week

Dashboards with filters people can change themselves, organized by team, and sent to the people who never log in.

- ### Find the right dashboard

  Collections keep each team's dashboards together, and verified dashboards, metrics and questions rank higher in search.

- ### Dashboards, Documents or data apps

  Dashboards for the numbers people check every week, Documents for an analysis with its charts in it, and data apps (beta) when a team needs its own tool.

- ### Reports that come to people

  Send a dashboard by email or Slack on a schedule, even to people without a Metabase account, and set alerts on the numbers you watch.

[Explore dashboards and reporting](https://www.metabase.com/features/analytics-dashboards)

## People, and their agents, only see what they should

Open data up to the whole company without opening up everything.

- **Groups, synced from your SSO.** Decide which data and collections each group can see, and who can edit.
- **Row and column security.** Sales sees its own accounts, finance sees every row, and nobody sees the columns you've hidden. See [permissions](https://www.metabase.com/features/permissions).
- **The same rules for AI.** Metabot and agents connected through the MCP server only see what the person using them can see. You choose which groups get AI, and which MCP tools each group's agents can use.
- **See everything that happens.** [Usage analytics and auditing](https://www.metabase.com/features/usage-analytics) show who viewed, ran, downloaded or changed what across your Metabase, including every Metabot conversation and every call an agent makes through the MCP server.

[Explore data segregation](https://www.metabase.com/features/data-segregation)

## Your team deserves better than vibes

Point Claude or ChatGPT straight at your warehouse and it signs in with a shared credential, guesses at your metrics, and leaves the answer in a chat. With Metabase, your team keeps its AI and gets the same governed numbers as everyone else.

|                            | Agent on your warehouse                       | Metabase                                            |
| -------------------------- | --------------------------------------------- | --------------------------------------------------- |
| Sign-in                    | One shared database credential                | Each person signs in, with SSO                      |
| Permissions                | Every table the credential can read           | Each person's permissions, down to rows and columns |
| Numbers match your metrics | Definitions guessed from the schema, per chat | Metrics your data team defined once                 |
| Where answers live         | A chat window, until it scrolls away          | Dashboards and subscriptions the whole team uses    |
| Audit                      | Database query logs, if you kept them         | Who asked what, and which data it touched           |

## Run it where your data has to stay

Metabase queries your database where it is and doesn't store a copy. Run it on our cloud or yours, and connect [20+ data sources](https://www.metabase.com/data-sources).

- ### Metabase Cloud

  We run it for you, in the US, the EU, South America or Asia Pacific.

- ### Self-hosted

  Run it on your own servers, with AI on the model you choose, including one you host yourself.

- ### Open source

  The core of Metabase is open source and free to run. See the code, and start without a contract.

> “Thanks to Metabase I don’t need to ask an engineer to get me some special data from our customer database. I can make that query on my own and help out the customer in minutes vs hours or days.”

Peer Richelsen Co-founder, Cal.com

#### Frequently asked questions

What is self-service business intelligence?

Self-service BI means people outside the data team can answer their own questions with data, without filing a request. In Metabase, the data team connects your database, sets permissions and defines the metrics that matter. Everyone else explores with the [query builder](https://www.metabase.com/features/query-builder) and drill-through, asks [Metabot](https://www.metabase.com/features/metabase-ai) in plain English, or asks their own AI client through the [MCP server](https://www.metabase.com/features/mcp-server), and every route uses the same definitions and permissions.

Do we still need a BI tool if our team already uses Claude or ChatGPT?

Your team can keep using Claude or ChatGPT. Connected to your database directly, an agent uses one shared credential, guesses at your metrics from the schema, and leaves its answers in a chat. Connected to Metabase through the MCP server, it signs in as the person asking, sees only what they can see, favors the metrics and models you've published, and can save what it builds as a dashboard the rest of the team can open. Metabase is also where the dashboards people check every week live, and where you see who asked what.

Can people use Metabase without knowing SQL?

Yes. In the [query builder](https://www.metabase.com/features/query-builder), people pick data, filter and summarize without SQL, and drill-through lets them click any chart to break it out or see the rows behind it. They can also ask Metabot in plain English. Analysts who prefer SQL can use the [SQL editor](https://www.metabase.com/features/sql-editor).

How do we keep AI answers consistent with our metrics?

Your data team defines metrics, table descriptions and business terms once, in your [semantic layer](https://www.metabase.com/product/data-studio). Metabot and agents connected through the MCP server favor those definitions when they search, and Metabot shows the query behind every answer, so anyone can check how a number was produced. As with any AI, people should still check answers that matter. See [Metabot](https://www.metabase.com/docs/latest/ai/metabot).

Can each person see only their own data?

Yes. Data permissions decide which databases and tables each group can query, and [row and column security](https://www.metabase.com/docs/latest/permissions/row-and-column-security) filters rows and hides columns based on who's signed in. The same permissions apply to Metabot and to agents connected through the MCP server. See [permissions](https://www.metabase.com/features/permissions).

Does Metabase need a data warehouse?

No. Metabase connects to the database or warehouse you already have, including Postgres, MySQL, Snowflake, BigQuery, Redshift and Databricks, and queries it directly, so reports show live data. It doesn't copy or store your data. You can also [upload CSV files](https://www.metabase.com/product/csv-uploads). See all [data sources](https://www.metabase.com/data-sources).

Does Metabase work with dbt?

Yes. Keep dbt to build your tables. Metabase connects to the tables dbt builds, and your data team defines the metrics and descriptions people and AI use on top of them in Metabase's [semantic layer](https://www.metabase.com/product/data-studio).

Can we move from the BI tool we use now?

Yes, though there's no automatic import, so you rebuild the dashboards people still use. A coding agent can do much of that work through the [Metabase CLI](https://www.metabase.com/product/analytics-as-code) or the MCP server, with your team reviewing what it builds.

Is Metabase open source, and can we self-host it?

Yes. The core of Metabase is open source, and you can run it on your own servers for free, including Metabot on your own AI provider and the MCP server. Pro and Enterprise also run self-hosted, and Enterprise can run fully air-gapped. Or let us host it on [Metabase Cloud](https://www.metabase.com/cloud).

Which plans include these features?

The query builder, SQL editor, dashboards, Documents, subscriptions, alerts, Metabot (on your own AI provider) and the MCP server are on every plan, including open source. Pro and Enterprise add row and column security, SSO (SAML, JWT and OIDC), content verification, the Library, dependency checks, data apps (beta), AI access controls and usage analytics. Enterprise adds air-gapped deployment, procurement support and a dedicated success engineer. See [pricing](https://www.metabase.com/pricing), or [talk to sales](https://www.metabase.com/sales).

### Every team on the same numbers, without the request queue

Start a 14-day free trial of Metabase Cloud, or install open source and run it yourself. Row and column security, SSO and AI controls are on Pro and Enterprise. Need a contract or help with procurement? [Talk to sales](https://www.metabase.com/sales).

[Get started](https://store.metabase.com/checkout) [Install open source](https://www.metabase.com/start/oss)
