Metabase Integrations: What It Connects To — and What You Still Have to Build
Someone on your team needs to connect Stripe data to a dashboard. Or HubSpot pipeline numbers to a board report. Or Shopify orders to something the ops lead can actually query. Metabase looks like the right tool — it's free, well-regarded, and the setup docs make it sound straightforward.
Then you open the connectors page and find databases: PostgreSQL, MySQL, Snowflake, BigQuery. Not Stripe. Not HubSpot. Not any of the SaaS tools that actually run your business.
That's not a gap in the documentation. Metabase is a visualization layer — it connects to a database you already have and lets people explore that data without writing SQL. It does this well, and thousands of teams use it for exactly that. But it doesn't store your data, move your data, or pull it from your SaaS tools. Those are separate problems, and for most teams setting up analytics for the first time, they're the ones that turn a BI project into a data infrastructure project.
Here's what Metabase actually reads, what you still have to build, and what that costs.
In short: Metabase has 18 official database drivers — all databases, zero SaaS connectors. To get data from Stripe, HubSpot, Salesforce, or any SaaS tool into Metabase, you need to build a pipeline (ETL), set up a warehouse, and maintain both. The "free and open-source" BI tool assumes a data stack that typically costs $1,200–$8,700/month depending on your warehouse choice and how much engineering time you're putting in.
What Metabase Actually Connects To
Metabase's data sources page lists 20 official data sources. Eighteen of those are database drivers maintained by the Metabase team:
Athena, BigQuery, ClickHouse, Databricks, Druid, MariaDB, MongoDB, MySQL, Oracle, PostgreSQL, Presto, Redshift, Snowflake, SparkSQL, SQL Server, SQLite, Starburst, and Vertica.
The other two — Google Sheets and CSV uploads — are Cloud Storage features available only on Metabase Cloud (not self-hosted), with their own pricing starting at $40 per 500,000 rows.
Another ~10 databases are supported through community drivers — DuckDB, Exasol, Firebolt, and others. These are self-host only, maintained by third parties, and not officially supported.
Every one of these connections assumes the same thing: your data is already in a database. Metabase reads databases. It doesn't fill them.
What "supported" actually means
"Metabase supports Snowflake" means Metabase can connect to a Snowflake instance you've already provisioned, loaded with data, and modeled. It doesn't mean Metabase will get your Stripe charges into Snowflake, set up the tables, or keep the sync running. That's a separate stack — and usually the expensive part.
Metabase also has transforms — SQL and Python operations that run on data already inside your database. These can reshape, aggregate, and join tables that already exist, similar to what dbt does but more limited (one transform at a time, in-memory, pandas/numpy only for Python). They're useful once your data is in the database — but they can't get it there. The E and L in ETL are missing; only the T is there.
Connecting Metabase to Specific Sources
The question is almost always the same: "can I connect Metabase to [thing]?" Here's what that actually looks like for the most common sources.
Google Sheets
On Metabase Cloud, Google Sheets syncing is available through the Cloud Storage add-on — $40 per 500,000 rows, 15-minute sync. Self-hosted Metabase can't connect to Sheets at all. Even on Cloud, it's narrow: ClickHouse under the hood, 50MB CSV limit. It handles spreadsheet data, not SaaS API volume.
Snowflake
Metabase connects to Snowflake natively — it's one of the 18 official drivers. The connection itself works well. The question is whether your data is in Snowflake yet. If you're running Stripe, HubSpot, and Google Analytics, you need an ETL tool (Fivetran, Airbyte, or alternatives) to move that data into Snowflake before Metabase can read it. Snowflake alone runs $500–$1,500/month at startup scale, before the ETL tool, before Metabase.
For more on Snowflake costs, see our Snowflake pricing breakdown.
BigQuery
Also natively supported. BigQuery's on-demand pricing and 1TB free analysis tier make it significantly cheaper than Snowflake at startup scale — often $50–$300/month. But the same pipeline question applies: you still need Fivetran, Airbyte, or custom scripts to get your SaaS data into BigQuery tables before Metabase shows a chart.
ClickHouse
Official, core driver since Metabase 54 (April 2025). Bundled with Metabase and available on Cloud. If you're already running ClickHouse, the connection is straightforward.
DuckDB
Community driver maintained by MotherDuck. Self-host only — not available on Metabase Cloud, not officially supported. If you're evaluating DuckDB, the Metabase connection requires self-hosting with no official maintenance guarantees.
PostgreSQL
The simplest path. If your application data already lives in Postgres, Metabase connects to it directly. Many teams start here — point Metabase at your production database and start querying. It works until analytics queries start competing with application queries for resources. That's when you hit the Postgres analytics wall and need a separate analytics warehouse, a read replica, or a platform where the analytics engine scales without replatforming.
SaaS Sources (Stripe, HubSpot, Salesforce, Shopify)
This is where the "integration" story breaks. Metabase has zero drivers for SaaS applications. No Stripe connector, no HubSpot connector, no Salesforce, no Shopify, no Google Analytics.
To see your Stripe data in Metabase, you need to:
- Set up an ETL tool (Fivetran, Airbyte, or custom code) to pull data from Stripe's API
- Load that data into a warehouse Metabase can read (Snowflake, BigQuery, Postgres)
- Model the raw data into queryable tables (dbt, manual SQL, or the warehouse's own tooling)
- Connect Metabase to the warehouse
- Maintain the pipeline — monitor syncs, handle API changes, manage schema drift
Multiply that by every SaaS source you need. Each one is another pipeline to build, another sync to monitor, another thing that breaks at 2 AM.
For a walkthrough of what this looks like for Stripe specifically, see Sync Stripe to a Data Warehouse.
What the Full Stack Actually Costs
The Metabase pricing page is clear about what Metabase charges: Open Source is free, Starter is $100/month for 5 users, Pro is $575/month for 10 users, Enterprise starts around $20,000/year. Cloud plans also carry usage-based charges — AI tokens at $3.75 per million, transforms at $0.01–$0.02 per run, and storage starting at $40 per 500K rows.
But Metabase's price is rarely the expensive line. Here's the stack most teams end up building underneath it:
| Component | Typical monthly cost |
|---|---|
| Metabase (self-hosted OSS) | $0 |
| Self-hosting infrastructure | $100–$200 |
| ETL tool (Fivetran or Airbyte) | $0 (free tier)–$2,000 |
| Analytics warehouse | $50–$1,500 (BigQuery) or $500–$1,500 (Snowflake) |
| Engineering time (10–20 hrs/week) | $2,000–$4,000 |
| Total | $1,200–$8,700/month |
The self-hosting estimate comes from Metabase's own documentation: HA servers ~$48, load balancer ~$12, managed database ~$40–$60, SMTP ~$12.
Fivetran has a free tier covering up to 500,000 monthly active rows — enough for very early-stage teams with minimal data volume. Beyond that, costs scale with connectors and data volume. A startup with 5–10 connectors and moderate volume typically pays a few hundred to $2,000/month.
The engineering time is the swing factor. Teams with a dedicated data engineer maintaining the pipeline, monitoring syncs, and managing schema changes are at the high end. Teams without dedicated data staff are either at the low end (patching things reactively) or about to discover they need someone.
For the fully-loaded cost breakdown by company stage, see our detailed pricing analysis.
When Metabase's Architecture Makes Sense
Metabase's architecture isn't wrong — it's just designed for a specific situation.
You already have the stack. If a data engineering team maintains your warehouse, runs your pipelines, and manages your models, Metabase is a solid visualization layer on top. The infrastructure cost is already sunk. You're adding a BI seat, not a data project.
You only query one database. If everything lives in a single Postgres instance and that's genuinely enough, Metabase pointed at it is simple, close to free, and effective. You don't need pipelines because your data is already where it needs to be.
You need embedded analytics. Metabase Pro and Enterprise have mature embedding and white-labeling for customer-facing dashboards. That specialization has real value.
When it doesn't
The architecture stops making sense when you're the one building the stack underneath. When you need data from 3–5 SaaS tools, don't have a warehouse yet, and don't have a data engineer — you're not evaluating a BI tool anymore. You're scoping a data platform project.
One fintech team we talked to had built exactly this path: custom AWS Glue jobs syncing four databases plus Stripe into Redshift every 12 hours, with Metabase on top. The CEO was still doing board reporting in Google Sheets because the dashboards couldn't produce the views investors needed. The BI tool was free. The infrastructure project around it was not.
Another company ran Metabase alongside dbt, Redshift, and Tableau for years, paying a dedicated consultant just to keep dashboards in sync with the constantly shifting dbt models underneath. When they consolidated to a single platform, the consultant's role disappeared within months — not because the work got automated, but because the architecture that generated it was gone. Assembling six tools means maintaining six tools.
What Changes with a Platform
Metabase reads databases. Definite is the database, the pipelines, and the analytics in one system.
| Metabase | Definite | |
|---|---|---|
| SaaS connectors | None — Stripe, HubSpot, Shopify require separate ETL | Built-in — no pipeline to build |
| Warehouse | Bring your own (Snowflake, BigQuery, Postgres) | Built-in, managed |
| Data modeling | External (dbt, manual SQL) | Governed semantic layer, included |
| AI analyst | Metabot add-on ($100/mo, 500 requests) | Fi, included (credit-based) |
| Pipeline maintenance | Your team's problem | Managed — no syncs to monitor |
| Pricing | Free–$575/mo + the stack underneath | $250/mo flat, unlimited users |
The difference isn't the number of connectors. It's that the connector question disappears — ingestion, storage, modeling, and BI are one system. You connect Stripe, HubSpot, or Shopify and start querying. There's no pipeline to build, no warehouse to provision, no engineer keeping syncs alive at 2 AM. Fi doesn't just answer questions about your data — it builds the dashboards, writes the SQL, and modifies the models. And technical teams still get full SQL, Python, and MCP access inside the governance boundary.
The tradeoff is real: you're choosing one system instead of assembling your own, which means you're betting on that system scaling with you. For most teams without a data engineering org, that bet is simpler than maintaining the alternative.
For a full comparison, see Definite vs Metabase.
Frequently Asked Questions
Does Metabase have ETL?
No. Metabase has no native data ingestion — it cannot pull data from Stripe, HubSpot, Salesforce, or any external SaaS source. It does have SQL and Python transforms that operate on data already in your database — useful for reshaping and aggregating, but these are in-database transformations, not data extraction. To get data into a database Metabase can read, you need a separate ETL/ELT tool.
What databases does Metabase support?
Metabase has 18 official database drivers: Athena, BigQuery, ClickHouse, Databricks, Druid, MariaDB, MongoDB, MySQL, Oracle, PostgreSQL, Presto, Redshift, Snowflake, SparkSQL, SQL Server, SQLite, Starburst, and Vertica. Google Sheets and CSV uploads are available on Metabase Cloud only, via the Cloud Storage add-on. Community drivers add DuckDB, Exasol, and others for self-hosted instances.
Can Metabase connect to Google Sheets?
On Metabase Cloud, yes — through the Cloud Storage add-on, which starts at $40 per 500,000 rows with 15-minute sync intervals. Self-hosted Metabase cannot connect to Google Sheets natively.
Can Metabase connect to Stripe or HubSpot?
Not directly. Metabase has no SaaS connectors. To see Stripe or HubSpot data in Metabase, you need an ETL tool (like Fivetran or Airbyte) to move the data into a database first, then connect Metabase to that database. See our guide on syncing Stripe to a warehouse.
Does Metabase have AI?
Metabase offers Metabot, an AI assistant available as a $100/month add-on on paid Cloud plans (500 requests included). It can answer questions about your data in natural language. Metabase also recently introduced an Agent API for building headless AI applications. For more on how AI works on Metabase versus platforms with built-in AI, see our AI comparison.