Best BI tools for Supabase (2026): dashboards that respect RLS, from AI to open source

Supabase gives you a Postgres database with row-level security, auth, and an auto-generated API, but no reporting layer. When a founder or PM asks for a dashboard, you need a BI tool that connects directly to Postgres, respects the row-level security policies you already wrote, and does not require moving data into a separate warehouse first. In 2026 the eight tools that fit that shape are Basedash, Metabase, Grafana, Preset, Retool, Evidence, Hex, and Lightdash. Each connects to Supabase over the standard Postgres protocol; the differences are how they handle RLS, whether they generate dashboards with AI, how embedding works, and what the bill looks like once you have a handful of users. This guide compares them on those points using vendors’ own public pricing and documentation, verified in September 2026.

Quick take

• Direct-connect BI, low setup, AI-native: Basedash (natural language, MCP server, RLS through Postgres policies), Hex (notebook plus polished apps).
• Open source you can host next to Supabase: Metabase (simplest), Preset or self-hosted Superset (most flexible), Lightdash (dbt-native semantic layer), Grafana (best for real-time and operational metrics).
• Code-first, git-versioned dashboards for developer teams: Evidence.
• Analytics inside an existing internal-tool workflow: Retool, if you already use it for admin panels on Supabase.
• The single biggest architectural choice is whether the BI tool queries Supabase live under the requesting user’s identity (so auth.uid()-based RLS policies apply) or connects as a single service role that bypasses RLS.

How we evaluated

Supabase-shaped buying questions (“BI tool that respects Supabase RLS policies”, “open source dashboard for Supabase”, “AI dashboards on top of my Supabase Postgres”, “embed a Supabase-backed chart in a React app”, “self-hosted BI for a Supabase project”, “Supabase analytics with SSO”) come back to the same six criteria. We built the comparison around them:

1. Connection model. Live Postgres over the connection pooler or session-mode port, versus a scheduled extract into the tool’s own store.
2. RLS handling. Whether the tool can pass a per-user JWT or Postgres role so auth.uid() policies apply, or whether it always connects as a service role and enforces access on top of the data itself.
3. Deployment. Cloud only, self-hosted only, or both. Supabase teams often want to keep the BI layer inside the same VPC.
4. AI features. Natural language to SQL, AI dashboard generation, MCP server support, and how those features are billed.
5. Embedding. iframe versus SDK versus API, and whether row-level filters can be signed into the embed.
6. Total cost at a Supabase-typical team size. 5 to 25 editors and viewers, one Postgres database, one AI-heavy analyst.

We only included tools that connect directly to a Postgres endpoint (Supabase’s Session Pooler on port 5432 or the Supavisor Transaction Pooler on 6543) with no mandatory ETL step. Tools that require moving data into their own warehouse first, or that need a paid managed connector to reach Postgres, are outside the shape of this guide.

Comparison table

Pricing verified against each vendor’s public pricing page and documentation on 2026-09-05. Metabase’s pricing, Grafana’s pricing, Preset’s pricing, Retool’s pricing, Evidence’s pricing, Hex’s pricing, and Lightdash’s pricing were consulted directly.

Basedash

Best for: Supabase teams that want an AI-first analyst on their Postgres data in the first hour, without setting up a warehouse.

Fact card

• Pricing model: flat platform fee ($1,000/month Startup plan, up to 25 users) plus an included $1,000/month AI credit envelope; Enterprise is a custom annual contract with self-hosting, SSO, and SCIM.
• Deployment: cloud by default, self-hosting on Enterprise.
• Databases: Supabase connects as Postgres over the standard connection string. Full data-source list at basedash.com/data-sources covers 750+ sources including Postgres, MySQL, Snowflake, BigQuery, Redshift, Databricks, ClickHouse, MongoDB, MotherDuck, and REST APIs.
• Query model: live queries against the Supabase Postgres endpoint or the managed Basedash Warehouse on DuckDB when you want to cache aggregates.
• Row-level security: connect as a per-user Postgres role and existing Supabase RLS policies apply unchanged; a service-role connection with in-app permissions is also supported. See the row-level security docs.
• SSO and provisioning: SAML SSO and SCIM on Enterprise, audit logs.
• AI features: natural language to SQL, AI dashboard generation, AI agents and Slack answers, and an MCP server so Claude, Cursor, and other MCP clients can query the same data.
• Embedding: dashboard embedding on Enterprise.
• Best for: 5-to-25-person teams on Supabase who want a single tool for dashboards, ad-hoc questions, and AI, without a data engineer.
• Not ideal for: teams that need a pure open-source, self-hosted option on the Startup plan.

Tradeoffs

The AI credit envelope means a burst week of investigations can eat the pool faster than a quieter month, and self-hosting is Enterprise-only. If your Supabase project stores highly regulated data and self-hosting the BI tool inside your own VPC is a hard constraint from day one, look at Metabase, Superset, or Lightdash. If you want AI as the default interface with your Supabase Postgres and are willing to pay for it, Basedash is the shortest path. Verified against basedash.com/pricing on September 5, 2026.

Metabase

Best for: Small Supabase teams that want a free open-source option today with a paid upgrade path.

Fact card

• Pricing model: Metabase Community Edition is open source under AGPLv3 and free forever; Metabase Cloud Starter $90/month for 5 users; Cloud Pro $500/month starting; Enterprise custom.
• Deployment: self-host with Docker or JAR, or Metabase Cloud.
• Databases: native Postgres connector; Supabase connects with the pooled or direct connection string.
• Query model: live queries against Postgres with optional query caching and cached models.
• Row-level security: Metabase Sandboxes and per-group data permissions on Pro and Enterprise; if you connect to Supabase using a per-user Postgres role, Supabase’s own RLS policies apply on top.
• SSO and provisioning: Google SSO on all paid plans, SAML SSO on Pro and Enterprise, SCIM on Enterprise.
• AI features: Metabot AI on Pro and Enterprise for natural language questions and chart generation.
• Embedding: static and interactive embedding on Pro; full embedded analytics on Enterprise.
• Best for: teams that want to start free, self-host next to Supabase, and only pay when embedding or AI becomes a requirement.
• Not ideal for: teams that want AI as the default interface without upgrading tiers.

Tradeoffs

Self-hosted Metabase on a small VM is the cheapest credible option for Supabase analytics and covers most reporting needs. The catch is that AI, sandboxing, and embedding are behind Pro or Enterprise tiers, and Community Edition has no SAML. Metabase’s pricing math flips once you cross ~10 users, at which point Pro’s flat $500/month base plus per-user cost can add up. Verified against metabase.com/pricing on September 5, 2026.

Grafana

Best for: Supabase teams whose primary need is operational and real-time dashboards on Postgres, not weekly business reports.

Fact card

• Pricing model: Grafana OSS is Apache 2.0 and free; Grafana Cloud has a free tier and a Pro tier starting from about $19/month plus usage-based charges for metrics, logs, and traces; Advanced and Enterprise are custom.
• Deployment: self-host anywhere or Grafana Cloud.
• Databases: PostgreSQL datasource ships in core Grafana and connects to Supabase with any Postgres connection string.
• Query model: live SQL queries with per-panel refresh intervals; optimized for time-series.
• Row-level security: Grafana does not enforce RLS itself. If you configure the Postgres datasource with per-user credentials you can push access down to Supabase RLS, but the default is a shared service credential.
• SSO and provisioning: SAML on Grafana Enterprise and Cloud Advanced.
• AI features: Grafana LLM app and Grafana ML for forecasting and anomaly detection.
• Embedding: public dashboards, iframe embedding, and scripted rendering.
• Best for: teams already using Grafana for infrastructure metrics who want to graph their Supabase Postgres tables in the same place.
• Not ideal for: teams that need governed metrics, self-serve business users, or interactive drill-down.

Tradeoffs

Grafana is the strongest tool for making a “how many signups per minute” Supabase panel that refreshes every 10 seconds, and it is free to self-host. It is the weakest of the eight tools on the parts of BI that involve non-technical users, semantic modeling, and governed metric definitions. Verified against grafana.com/pricing on September 5, 2026.

Preset (managed Superset)

Best for: Supabase teams that want Apache Superset’s flexibility without running it themselves.

Fact card

• Pricing model: Apache Superset is open source under Apache 2.0 and free; Preset Cloud has a free workspace, Professional at $20/user/month, and Enterprise custom.
• Deployment: self-host Superset with Docker or Helm, or use Preset’s managed cloud.
• Databases: Superset’s PostgreSQL driver connects to Supabase directly.
• Query model: live queries against Postgres with optional result caching in Redis.
• Row-level security: Superset defines RLS rules per role that append WHERE clauses to queries; if you connect per user, Supabase’s own RLS also applies.
• SSO and provisioning: SAML and OpenID Connect on Preset Enterprise.
• AI features: available through community plugins on Superset; not a first-class product feature on Preset.
• Embedding: Embedded SDK for signed embedded dashboards on Preset Enterprise.
• Best for: teams that want the widest chart library, deep SQL Lab, and a Python plugin surface.
• Not ideal for: teams that want AI as a headline capability or want the smallest possible setup.

Tradeoffs

Superset has the broadest visualization catalog of any open-source BI tool and connects to Supabase without friction, but its learning curve is real, and its native permissions model is more complex than Metabase’s. Preset takes the operational burden off but pushes SSO and embedding to Enterprise. Verified against preset.io/pricing on September 5, 2026.

Retool

Best for: Teams already building internal tools on Supabase in Retool who want simple charts and reports without introducing a second product.

Fact card

• Pricing model: Free for up to 5 users; Team at $10 per standard user/month plus $50 per end user/month; Business and Enterprise custom.
• Deployment: cloud or self-host.
• Databases: first-class Supabase resource type plus generic Postgres; queries can run under a per-user JWT.
• Query model: live queries per component or workflow.
• Row-level security: the Supabase resource can pass the authenticated user’s JWT into the query, so Supabase’s RLS policies apply the same way they do from your app.
• SSO and provisioning: Google and Microsoft SSO on Team, SAML and SCIM on Business and Enterprise.
• AI features: Retool AI actions (LLM calls as workflow steps) and AI SQL for generating queries.
• Embedding: iframe embedding of Retool apps and Retool Portals for external users on Business and Enterprise.
• Best for: teams that already use Retool for admin panels on Supabase and want reporting inside the same tool.
• Not ideal for: teams that want a purpose-built BI experience with a semantic layer, or that need governed metric definitions.

Tradeoffs

Retool’s Supabase integration is genuinely per-user, which makes RLS enforcement clean. The tradeoff is that Retool is an app builder first and a BI tool second: dashboard interactions like cross-filtering, drill-down, and metric sharing are things you build yourself, not defaults. End-user pricing also adds up quickly if you need to publish dashboards to an audience of dozens. Verified against retool.com/pricing on September 5, 2026.

Evidence

Best for: Developer teams on Supabase who want dashboards that live in the same git repository as their app.

Fact card

• Pricing model: Evidence is open source under MIT and free to self-host; Evidence Cloud has a free plan and Team at $2,500/month; Enterprise custom.
• Deployment: self-host as a static site, or Evidence Cloud.
• Databases: PostgreSQL connector ships in Evidence; Supabase connects with the standard connection string.
• Query model: SQL queries in Markdown files that Evidence executes at build time or on demand in Cloud.
• Row-level security: not enforced at the Evidence layer. Because Evidence produces a static site, RLS enforcement has to happen when the SQL runs; per-user filtering is more work than in the other tools here.
• SSO and provisioning: SAML SSO on Evidence Cloud Enterprise.
• AI features: AI copilot in Evidence Cloud for query and page authoring.
• Embedding: publishes a static site you can host anywhere, or embed pages behind your own auth.
• Best for: engineering-led teams that want dashboards as code, reviewable in pull requests.
• Not ideal for: teams whose primary audience is non-technical users who want to build their own dashboards.

Tradeoffs

Evidence turns a Supabase-backed dashboard into a folder of Markdown and SQL that ships through the same CI as your app. If you are a five-engineer startup, that is a very clean fit. If you also need business users to explore data on their own, Evidence alone will not cover it; teams often pair it with Metabase or Basedash for the ad-hoc side. Verified against evidence.dev/pricing on September 5, 2026.

Hex

Best for: Analyst-led Supabase teams that want notebooks for exploration and polished apps for sharing.

Fact card

• Pricing model: Community free plan; Professional $36 per editor/month; Team $75 per editor/month; Enterprise custom.
• Deployment: multi-tenant cloud on standard plans; single-tenant cloud and VPC on Enterprise.
• Databases: PostgreSQL connector; Supabase connects with any Postgres connection string.
• Query model: SQL and Python cells run in Hex’s compute; results feed downstream cells and charts.
• Row-level security: Hex projects have per-project and per-app permissions; Supabase RLS applies if you connect with a per-user or dynamic Postgres identity.
• SSO and provisioning: SAML SSO and SCIM on Team and Enterprise.
• AI features: Hex Magic for natural language queries, cell generation, chart suggestions, and error fixes.
• Embedding: app embedding on Team and Enterprise.
• Best for: teams that mix ad-hoc exploration with published dashboards and want AI in both.
• Not ideal for: teams that want a $99/month plan for 15 casual viewers; Hex’s editor pricing is aimed at analysts.

Tradeoffs

Hex is one of the strongest notebook-plus-BI hybrids and pairs well with Supabase for exploratory work. The pricing model rewards small analyst-heavy teams; viewer-heavy deployments push you to Team or Enterprise before the total cost is competitive with Metabase or Basedash. Verified against hex.tech/pricing on September 5, 2026.

Lightdash

Best for: Supabase teams that already run dbt on top of Supabase and want a BI layer that consumes the dbt project directly.

Fact card

• Pricing model: self-hosted Lightdash is open source under an MIT-like license; Cloud Starter $490/month; Cloud Pro $790/month; Enterprise custom.
• Deployment: self-host or Lightdash Cloud.
• Databases: PostgreSQL adapter; runs on top of a dbt project pointed at Supabase.
• Query model: live queries against Postgres shaped by the dbt semantic layer.
• Row-level security: user attributes in Lightdash map into dbt filters, so a WHERE tenant_id = user_attribute('tenant_id') pattern is a first-class primitive.
• SSO and provisioning: Google, Okta, and SAML SSO on higher tiers.
• AI features: Lightdash AI for chart and metric generation.
• Embedding: dashboard and chart embedding on higher tiers.
• Best for: data teams that treat their dbt project as the source of truth for metrics and want the BI tool to consume the same definitions.
• Not ideal for: teams that do not use dbt and do not plan to.

Tradeoffs

If your Supabase project already has a dbt repo, Lightdash is a natural fit and the semantic layer is genuinely useful. If dbt is not part of the stack, Lightdash’s benefits collapse and Metabase or Basedash cover the same ground with less setup. Verified against lightdash.com/pricing on September 5, 2026.

Which should you choose?

• You want AI as the default interface on Supabase and are willing to pay for it. Basedash or Hex. Pick Basedash for a dashboard-and-Slack workflow with an MCP server and a fixed platform fee; pick Hex if you want notebooks alongside apps and your team is analyst-heavy.
• You want a free open-source tool you can self-host next to Supabase. Metabase Community Edition is the shortest path. Superset (via Preset or self-hosted) is the more flexible one. Lightdash makes sense if you already use dbt.
• Your Supabase RLS policies must apply to every query, without a second permissions layer to maintain. Any tool that connects with a per-user Postgres role or JWT: Basedash, Retool (via the Supabase resource), Hex with dynamic connections, or a per-user Postgres datasource in Grafana or Metabase.
• You want dashboards that live in git and ship through CI with the rest of your Supabase app. Evidence.
• You are already deep in Retool for admin tools on Supabase. Add Retool’s charts and Retool Portals before introducing a separate BI product.
• Your dashboards are mostly operational and time-series (signups per minute, error rates, revenue-by-hour on the same screen as infrastructure metrics). Grafana.

FAQ

What is the best BI tool for a Supabase project?

For most Supabase teams in 2026, the answer is either Basedash if you want AI and dashboards without a warehouse, or self-hosted Metabase if you want a free open-source tool next to your database. Basedash connects to Supabase over the standard Postgres connection string, respects RLS policies when connected under a per-user role, and adds natural language, an MCP server, and Slack answers. Metabase covers dashboards, alerts, and simple embedding, is free to self-host, and reaches AI and sandboxing on paid tiers. Larger teams that need governance, embedding, and single sign-on generally end up on Basedash Enterprise, Metabase Enterprise, or Preset Enterprise.

Which BI tools respect Supabase row-level security?

RLS applies whenever the BI tool queries Postgres under the identity RLS policies expect. In practice that means Basedash with per-user Postgres roles, Retool using the Supabase resource with an authenticated JWT, Hex with dynamic connections, and Metabase, Grafana, Superset, or Lightdash if you set up per-user Postgres datasources or user attributes that translate to the same filters. Any BI tool connected as a single service role bypasses RLS entirely; in that case, access has to be enforced inside the BI tool with sandboxes, row rules, or user attributes.

How do I connect a BI tool to Supabase?

Every tool in this guide connects to Supabase over the standard Postgres wire protocol. You use the Supabase Session Pooler on port 5432 for persistent connections (Metabase, Grafana, Preset, Lightdash) or the Supavisor Transaction Pooler on port 6543 for short-lived or serverless clients (Evidence build steps, Hex compute). Basedash, Retool, and Hex all treat Supabase as a Postgres source; Retool additionally has a first-class Supabase resource that can pass the user’s JWT. The connection details live under Project Settings, Database, Connection string in the Supabase dashboard.

Do I need a data warehouse to run BI on Supabase?

For most teams the answer is no. Supabase’s Postgres is fast enough to serve dashboards for hundreds of daily users if you index the tables the dashboards query, use materialized views for expensive aggregates, and run the BI tool through the Session Pooler. A warehouse becomes worth introducing when analytical queries start impacting transactional performance, when data volumes exceed a few hundred gigabytes, or when you need to join Supabase data with data from other systems. Until then, a live-query BI tool is cheaper and simpler.

Which open-source BI tool works best with Supabase?

Metabase for the fastest path from Docker to a dashboard, Superset for the widest chart library and most control, Lightdash if you already treat dbt as the source of truth, and Grafana if your dashboards are real-time and operational. All four connect to Supabase over standard Postgres and can be self-hosted next to it. Metabase is the easiest to operate for a small team; Superset scales further; Lightdash is the strongest choice when a semantic layer matters; Grafana is the most focused tool for time-series.

Can I embed a Supabase-backed dashboard in my own app?

Yes. Basedash supports dashboard embedding on Enterprise, Metabase has interactive embedding on Pro and full embedded analytics on Enterprise, Preset offers an Embedded SDK on Enterprise, Retool exposes apps through iframes and Portals, and Hex embeds apps on Team and Enterprise. Evidence produces a static site you can host anywhere. For customer-facing embedding where each customer should only see their own data, prefer tools that accept signed row-level filters (Metabase Enterprise, Preset Enterprise, Basedash Enterprise) or that let you push identity down to Supabase RLS (Retool, Hex with dynamic connections).

How much should a Supabase team budget for BI?

For a 10-person Supabase team, three realistic budgets in September 2026: self-host Metabase or Superset on a small VM for near zero if someone owns the operational work; Metabase Cloud Starter at $90/month for up to 5 users, then Pro at $500/month base once you outgrow that; Basedash Startup at $1,000/month for up to 25 users with a $1,000/month AI credit envelope. Hex and Lightdash Cloud start higher on a per-editor basis and are worth it when the analyst team is already investing in notebooks or dbt.

Related reading

• Best BI and dashboarding tools for PostgreSQL (2026): the broader Postgres comparison.
• Best BI tools for row-level security (2026): a deeper look at how BI tools implement RLS.
• Basedash pricing: the current Basedash plans and AI credit envelope.