Turn Your dbt Project Into a Semantic Layer for Agents

Draft placeholder. Full article coming soon.

You already have dbt models, tests, and a manifest. This post will show how to turn that into a governed semantic layer agents can query, without rebuilding your stack.

What this will cover

• Reading your dbt manifest to bootstrap metrics and relationships
• Defining metrics once so every consumer gets the same number
• Tagging the tenant key that scopes each customer’s rows
• Exposing governed metrics to agents over MCP
• Charting the agent’s query results in Claude or ChatGPT with @bonnard/mcp-charts

Frequently asked questions

Do I need to migrate off dbt?

No. A semantic layer reads your existing dbt models. You keep dbt as your transformation layer.

What if I only have models, not a semantic layer?

That’s the common case. You bootstrap from the manifest and define metrics on top.

How do agents chart the results?

Add a visualize tool to your MCP server with @bonnard/mcp-charts. The agent queries a metric, your runSql returns the rows, and an interactive chart renders in Claude or ChatGPT. Bonnard never touches your database. See the MCP Charts guide.