# dbt gives you model and column lineage inside the project, and it stops at the project boundary and the dashboard

GUIDE, read 10 Nov 2026

**The short answer.** dbt Catalog draws column lineage from the SQL of a dbt project, shows transformation formulas and carries descriptions down passthrough columns. Column lineage in Catalog needs an Enterprise plan, ignores joins and filters, and reaches BI only as dashboard exposures. OpenLineage, Monte Carlo, Collibra, Secoda and Euno each read dbt differently.

## dbt documents the project it runs

dbt's lineage is the lineage of a dbt project. dbt Catalog, per its docs, shows resources including models, tests and metrics with their lineage. dbt says it documents almost every project object through the description key, shown in Catalog. This is lineage with good provenance: the graph comes from the code that builds the tables, so the edge between two models is there because one selects from the other.

dbt's documentation now describes two options. dbt Docs v2 is an open-source site generated locally. dbt Catalog runs on the dbt platform and is updated in real time. The former Explorer is now documented as Catalog. If you read an older blog post that says Explorer, it describes the same product.

## Column lineage comes from parsing your SQL

dbt Labs' Catalog provides end-to-end column-level lineage for each column of a model, source or snapshot. The graph is built from SQL parsing of models. dbt says it can be incomplete when parsing fails, and that it reflects select statements only, not joins and filters. A column used only in a join key or a WHERE clause will not appear as feeding the output.

Two features go beyond a plain edge. The column evolution lens shows when a column is transformed versus passed through or renamed, and previews the transformation formula. Passthrough and renamed columns also inherit descriptions from upstream columns automatically, as far back as possible. For a team that documents sources once and wants that text to appear on every model, this saves real work.

With dbt v2, column lineage can also be generated locally, through the VS Code extension, dbt Docs v2 or the command line, under strict static analysis. That puts column lineage in front of a developer before a pull request exists.

## Column lineage depends on the plan

Catalog is available on dbt platform Starter, Enterprise and Enterprise+ plans. Column lineage in Catalog needs Enterprise or Enterprise+. dbt's pricing page lists Starter at $100 per user per month, including dbt Catalog lineage, and a free Developer plan. Starter includes 15,000 successful models built and 5,000 queried metrics per month.

So the table-level graph is available to a small team on Starter, and the column-level graph is available on the plans that start with a sales conversation. Read the dates in the sources: these figures were read on 10 Nov 2026.

## Where dbt lineage stops

dbt lineage covers dbt project resources. Its reach into BI is through exposures. dbt exposures record downstream dashboards in the project and appear in Catalog. dbt can also create downstream exposures automatically for supported integrations, and its docs name Tableau.

An exposure is a dashboard-level node. We found no dbt docs statement that Catalog column lineage extends into BI fields: none found on 10 Nov 2026. On the matrix, dbt is marked partial for BI field lineage, and partial for metrics. So the path from a Tableau field back to a dbt column is not drawn by dbt. Tools that document BI-field lineage, such as DataHub, Euno, Foundational and Select Star, draw it from the BI side.

The other boundary is the project itself. Data loaded by an ingestion tool before dbt, or consumed by a Python notebook after it, is outside the graph unless declared as a source or an exposure.

## Lineage refreshes when production runs

dbt updates Catalog lineage after each production or staging run, and v2 jobs refresh metadata on every run. dbt also says CI jobs do not update Catalog. Catalog therefore shows what was last run in production, not the state of a pull request. For impact analysis before merge, you need to compare your diff to Catalog yourself.

## How other tools read a dbt project

Many tools in this comparison start from dbt's artifacts or from the warehouse. They do not give the same answers. The table lists what each documents.

| Tool | How it reads dbt | Documented limit or note |
| --- | --- | --- |
| OpenLineage | Integration parses manifest.json, run_results.json and optionally catalog.json after a run; a structured log processor can extract lineage in real time | Column data types come from catalog.json |
| Monte Carlo | Takes table-to-table lineage from warehouse query logs, not from the dbt manifest, and overlays dbt job and run context | dbt context in lineage can take up to 24 hours |
| Collibra | dbt Core lineage supports 14 adapters, including Snowflake, BigQuery, Databricks and Redshift | dbt snapshots are not supported |
| Secoda | Pulls dbt Cloud compiled SQL, column descriptions and tags, tests and exposures; dbt v1.5 or less metrics become dictionary terms | Metric import is tied to dbt v1.5 or earlier |
| Euno | dbt Cloud integration retrieves artifacts from completed dbt runs automatically | Follows completed runs, not a manual export |
| Coalesce Catalog | GitHub Actions workflow syncs the dbt manifest on every push; manual lineage importer for gaps | Needs workflow set up |

## dbt's own MCP server

dbt's MCP server gives AI tools governed access to models, metrics, lineage and freshness. It comes self-hosted (uvx dbt-mcp) or remote on the dbt platform. dbt describes it as a read-only access layer that does not store production data. It also exposes text-to-SQL and SQL execution, so, unlike a context-only server, it can run queries. That is useful for an agent that checks a number after tracing it, and it is a point to review for access control.

## Checking dbt column lineage on your own project

Before you rely on column lineage from any tool, check three models in your own project. Choose one with a plain passthrough column, one with a calculated column, and one where a column is used only in a join condition. In dbt Catalog, open each column and read its lineage. The passthrough should show a clean edge and an inherited description. The calculated column should show the formula in the column evolution lens. The join-key column, by dbt's own description, will not show a downstream edge, which confirms the documented limit and shows you what it looks like in your project.

Then repeat the check in the tool you plan to use for the BI hop. If it reads dbt artifacts, check whether it sees the same model count as dbt does, since a mismatch usually means a stale artifact upload. If it reads the warehouse, check that it sees dbt-built views. The goal is a short list of cases where two tools disagree. Each disagreement tells you which tool to trust for which layer.

For the transformation layer, dbt has the best provenance of any tool here, since the graph is the code. Its weaker sides are the plan gating of column lineage, the select-only column edges and the dashboard-level BI reach. A common arrangement is dbt as the source of truth for models, plus another tool for the BI hop. See [the dbt page](/tools/dbt/) and [the matrix](/matrix/) for the evidence.

## Sources

- dbt: docs.getdbt.com/docs/explore/column-level-lineage: https://docs.getdbt.com/docs/explore/column-level-lineage (read 10 Nov 2026)
- dbt: docs.getdbt.com/docs/explore/explore-projects: https://docs.getdbt.com/docs/explore/explore-projects (read 10 Nov 2026)
- dbt: docs.getdbt.com/docs/build/documentation: https://docs.getdbt.com/docs/build/documentation (read 10 Nov 2026)
- dbt: docs.getdbt.com/docs/build/exposures: https://docs.getdbt.com/docs/build/exposures (read 10 Nov 2026)
- dbt: www.getdbt.com/pricing: https://www.getdbt.com/pricing (read 10 Nov 2026)
- dbt: docs.getdbt.com/docs/dbt-ai/about-mcp: https://docs.getdbt.com/docs/dbt-ai/about-mcp (read 10 Nov 2026)
- OpenLineage: openlineage.io/docs/integrations/dbt: https://openlineage.io/docs/integrations/dbt (read 10 Nov 2026)
- Monte Carlo: docs.getmontecarlo.com/docs/dbt-integration: https://docs.getmontecarlo.com/docs/dbt-integration (read 10 Nov 2026)
- Collibra: productresources.collibra.com/docs/collibra/latest/Content/CollibraDataLineage/D: https://productresources.collibra.com/docs/collibra/latest/Content/CollibraDataLineage/DataSources/dbtCore/ref_dbt-core-supported-transformations.htm (read 10 Nov 2026)
- Secoda: docs.secoda.co/integrations/data-transformation-tools/readme-1/dbt/metadata-extr: https://docs.secoda.co/integrations/data-transformation-tools/readme-1/dbt/metadata-extracted.md (read 10 Nov 2026)
- Euno: docs.euno.ai/sources/transformation-etl/dbt-cloud.md: https://docs.euno.ai/sources/transformation-etl/dbt-cloud.md (read 10 Nov 2026)
- Coalesce Catalog: docs.coalesce.io/llms-catalog.txt: https://docs.coalesce.io/llms-catalog.txt (read 10 Nov 2026)

Last reviewed 10 Nov 2026. Canonical: https://datalineagecompare.com/guides/dbt-lineage/
