Superset Alternative: Governed Analytics for B2B Products

Apache Superset is the most popular open-source BI tool after Metabase. It's powerful for SQL-native teams: connect to a warehouse, write SQL, build dashboards, share with your team. For internal analytics, it's hard to beat at the price (free).

But Superset was built for internal teams, not B2B products. Multi-tenancy is limited. There's no semantic layer. No AI agent support. Embedding requires Superset Server and custom auth. If you're shipping analytics to customers or connecting AI agents to governed data, Superset hits the same walls as Metabase.

Bonnard vs Superset at a glance

Feature	Superset	Bonnard
Primary use case	Internal SQL dashboards	Customer-facing analytics for B2B
Semantic layer	No	Yes (YAML cubes + views)
AI agent support	No	MCP server with publishable keys
Multi-tenancy	Limited (role-based, manual)	Built-in (security context, token exchange)
Embedded analytics	Iframe (requires Superset Server)	React SDK (native components)
Dashboards	Visual builder + SQL Lab	Markdown dashboards, CLI-deployed
Pre-aggregation / caching	Dashboard-level cache	Built-in pre-aggregation cache
Metric governance	No (ad-hoc SQL per chart)	YAML definitions, versioned in Git
CLI workflow	None	bon init, bon deploy, bon mcp, bon diff, bon schema
Self-host pricing	Free (Apache 2.0)	Free (Apache 2.0, all features)
License	Apache 2.0	Apache 2.0 (server), MIT (CLI)

Where Superset falls short for B2B

No semantic layer

Superset charts are SQL queries. Each chart defines its own metric calculation. Two dashboards can calculate "revenue" differently. There's no central definition. No governance. No single source of truth.

Bonnard defines metrics once in YAML. Every consumer (dashboards, React SDK, AI agents, APIs) queries the same governed definitions. See What Is a Semantic Layer?.

Multi-tenancy is manual

Superset has role-based access control, but it's designed for internal teams with different permission levels, not B2B products with hundreds of tenants. Isolating data per customer requires custom row-level security configurations per role or separate database connections per tenant. Neither scales.

Bonnard's security context enforces tenant isolation on every query structurally. Token exchange (bon_sk_... → scoped JWT) handles production multi-tenancy. No per-tenant role configuration.

No AI agent integration

Superset has no MCP support, no tool-use API, and no agent-compatible query interface. Getting data out of Superset into an AI agent requires hitting the Superset API, which returns dashboard data, not governed metrics.

Bonnard's MCP server lets any compatible agent discover and query governed metrics. Run bon mcp for connection configs.

Embedding is an iframe

Embedding Superset requires running Superset Server and configuring authentication (usually via a custom auth proxy). The embedded experience is an iframe. Styling is limited to Superset's theming options.

The @bonnard/react SDK provides native React components styled with your design system. No iframe. No Superset Server dependency.

Who should stay with Superset

Superset is the right tool if:

• SQL-native internal team that wants free, powerful dashboards
• No customer-facing or multi-tenant requirements
• You prefer visual SQL exploration (SQL Lab) over YAML configuration
• Your team is comfortable with Superset's deployment and maintenance

If you're evaluating Superset alongside other tools, see our comparisons with Metabase, Looker, and Tableau.

FAQ

Is Bonnard open source like Superset?

Yes. Both are Apache 2.0. Bonnard includes MCP, React SDK, multi-tenancy, pre-aggregation, and RBAC in the open-source version. No features are gated behind a paid tier.

Can I use Bonnard and Superset together?

Yes. Keep Superset for internal SQL exploration. Add Bonnard as the semantic layer for governed metrics, AI agent access, and customer-facing analytics. Both connect to the same warehouse.

Does Bonnard have SQL Lab?

No. Bonnard focuses on governed metric serving, not ad-hoc SQL exploration. For SQL exploration, use bon query from the CLI or connect an AI agent that can run sql_query through the semantic layer with access controls enforced.