Looker Alternative: Agent-Native Analytics with a Semantic Layer

Looker pioneered the semantic layer. LookML gave data teams a way to define business logic once and query it everywhere. That was a genuine step forward. But Looker is enterprise-only, expensive, and built for a world before AI agents existed.

Bonnard takes the same core idea, a governed semantic layer, and makes it open source, agent-native, and shippable from the terminal. Metrics defined in YAML, served via MCP to AI agents, embedded in your product with React components, and deployed with bon deploy. No LookML. No six-figure contract.

Bonnard vs Looker at a Glance

Feature	Looker	Bonnard
Semantic layer	LookML (proprietary)	YAML cubes + views
AI agent support (MCP)	No	Native (publishable keys per tenant)
Pricing	$60K+/yr (Enterprise only)	Free (Apache 2.0, all features)
License	Proprietary	Apache 2.0 (server), MIT (CLI)
Multi-tenancy	Row-level access per user group	Built-in (publishable keys + row-level security)
Embedded analytics	Looker Embed API + custom build	React SDK (BarChart, LineChart, BigValue)
Dashboards	LookML dashboards	Markdown dashboards, deployed via CLI
CLI workflow	None	bon init, bon deploy, bon mcp, bon query
Pre-aggregation / caching	Persistent Derived Tables	Built-in cache
Schema format	LookML	YAML
Self-hosting	Not available	Docker Compose, full features

Where does Looker fall short?

LookML Is Proprietary

LookML is a DSL that only works inside Looker. Your semantic layer is locked to a single vendor. If you leave Looker, you rewrite everything.

Bonnard uses YAML. Same cubes, measures, and dimensions concept, but in a format that lives in your repo, works with any editor, and isn't tied to a proprietary runtime. Version it in Git. Review it in PRs. No vendor lock-in.

No MCP Support

Looker has no MCP integration. You can query Looker's API from custom code, but there's no way for Claude, Cursor, ChatGPT, or CrewAI to connect directly to your Looker instance as a governed data source.

Bonnard deploys as an MCP server. bon mcp generates connection configs for your AI client of choice. For customer-facing use cases, generate publishable keys per tenant so your customers connect their own AI tools to their own data. Row-level security on every query.

Enterprise Pricing

Looker starts at roughly $60K per year and goes up from there. That's before you factor in Google Cloud platform costs, professional services, and the team needed to maintain LookML. For startups and mid-market companies, the math doesn't work.

Bonnard is Apache 2.0. Self-host for free with every feature included: MCP server, React SDK, markdown dashboards, multi-tenancy, pre-aggregation, RBAC, admin UI, CLI. Bonnard Cloud is available for teams that want managed infrastructure.

Embedding Requires Custom Build

Looker's embed SDK gives you iframes and an API. Building a polished embedded analytics experience means writing a custom frontend layer, handling auth flows, and managing Looker's embedding infrastructure.

The @bonnard/react SDK ships BarChart, LineChart, BigValue, and useBonnardQuery as production-ready components. Embed governed, multi-tenant charts directly in your product. No iframe, no custom auth layer.

No Markdown Dashboards

Looker dashboards are built in the Looker UI or defined in LookML. There's no way to author dashboards alongside your code and deploy them from the terminal.

Bonnard dashboards are written in markdown and deployed with bon deploy. Each tenant gets their own view, access-controlled automatically. Dashboards live in your repo next to your schema.

How does Bonnard differ?

YAML, Not LookML

Define total_revenue in a YAML cube. Query it from an AI agent via MCP, render it in a React BarChart, display it on a markdown dashboard, or fetch it from the TypeScript SDK. Same number everywhere. No proprietary DSL. What is a semantic layer?

Agent-Native from Day One

Bonnard was built for a world where AI agents query your data. MCP server, publishable keys per tenant, row-level security on every query. Your customers can connect Claude Desktop or Cursor to their own metrics without your team writing a single integration. Read more about agentic analytics.

Open Source, Full-Featured Self-Host

No tiered pricing. No enterprise-only features. The self-hosted version includes everything: MCP, React SDK, dashboards, multi-tenancy, pre-aggregation, RBAC, audit logging, admin UI, CLI. Apache 2.0.

CLI-First Workflow

bon init scaffolds your project. bon deploy pushes schema and dashboards. bon mcp configures AI agent connections. bon query tests from the terminal. bon datasource add --from-dbt imports existing dbt models. Ship from the terminal, not a GUI.

Who should stay with Looker?

Looker is the right tool if:

• You have a deep Google Cloud investment and Looker is already integrated with BigQuery, Vertex AI, and your GCP stack
• Your team has a large LookML codebase that would be costly to rewrite
• You need Looker's specific visualization capabilities and Explore interface for non-technical analysts
• Enterprise pricing isn't a constraint and you're satisfied with the existing Looker workflow

FAQ

Can I import my LookML into Bonnard?

Not directly. LookML is proprietary to Looker. But the concepts map cleanly: LookML views become Bonnard cubes, dimensions and measures translate to YAML equivalents. A typical migration takes days, not months.

Is the Bonnard server free to self-host?

Yes. Apache 2.0. The self-hosted version includes every feature: MCP server, React SDK, markdown dashboards, multi-tenancy, pre-aggregation, RBAC, admin UI, CLI.

Does Bonnard support BigQuery?

Yes. Same warehouse connectors as any semantic layer: Snowflake, BigQuery, PostgreSQL, Databricks, Redshift, DuckDB, ClickHouse. No data migration required.

How does Bonnard handle governance compared to Looker?

Bonnard defines access control in YAML alongside your schema. Row-level security, RBAC, and audit logging are built in. Every query, from every surface and tenant, resolves to the same governed metric definitions.

Can Bonnard work alongside Looker?

Yes. Some teams layer Bonnard on top of their existing warehouse alongside Looker. Bonnard handles the AI agent and embedded analytics use cases that Looker doesn't cover, while Looker continues serving internal Explore workflows.