dbt community feedback session: Agent Schema

​We're exploring a built-in way for dbt to write your project metadata to Agents Schema, so AI agents can query your models, columns, and dependencies. It would replace today's GitHub Actions setup. We'll show one proposed design and want you to find the holes.

​The idea in brief

• ​A flag such as --write-agents-schema controls when it writes. It works on compile, build, run, and docs generate, and you can set it in dbt_project.yml or turn it on in dbt Cloud job settings.

• ​An optional agents_schema: config controls where and what, including which folders, models, or tags to exclude.

• ​It writes the parsed project state, not run results, as a full replace on each write.

​We especially want to hear from you if you run dbt in production on dbt Cloud, Airflow, Dagster, CI, or something else. We want to confirm this works wherever you run dbt.

​Agenda

• ​Welcome and context (5 min)

• ​Demo (10 min)

• ​Proposed design (5 min)

• ​Discussion on triggers, config, and table shape (30 min)

• ​Wrap-up and next steps (5 min)

​Can't make it? Add your thoughts to the GitHub discussion: https://github.com/dbt-labs/dbt/discussions/15447
Background: /dbt-labs/agents_schema