In your dbt Cloud CI job configuration, add dbt sl validate as a step after dbt compile or dbt build to catch semantic layer errors on pull requests
The command runs three built-in validation passes: parsing (schema conformance), semantic (graph constraint checks), and data platform (verifying that referenced tables and columns exist in the warehouse)
Review the CI job logs for any validation failures, which will include the name of the failing semantic model or metric and the specific constraint violated
For local development, run dbt sl validate in your development environment with dbt Core and MetricFlow CLI installed to catch errors before pushing
Combine with dbt test and dbt unit test steps so that both transformation logic and semantic definitions are validated in the same CI run
Known gotchas
The data platform validation pass makes live warehouse calls; ensure your CI job's service token has at minimum SELECT access to the tables referenced in your semantic models
Parsing and semantic validations run without warehouse access and are fast; data platform validation is slower and can time out if your warehouse is cold or the manifest is large — consider running it only on changed files when possible
dbt sl validate does not validate saved queries exhaustively; test saved queries separately using dbt sl query --saved-query in a development environment
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp