Skip to main content

Optimization agent

The Optimization agent scans an existing dbt project for concrete, fixable issues and proposes targeted fixes - with cost and impact reporting so you can prioritize. This is the agent behind Optimize.

Input​

  • An existing dbt project (a connected repository and environment).

Output​

  • Candidates - individual detected issues.
  • Fixes - agent-generated changes for candidates you choose to fix.
  • Impact reports - the combined effect of a batch of fixes, including estimated cost change.
  • Pull requests for the fixes you apply.

What it looks for​

  • Performance & materialization - models that would benefit from a different materialization or a query rewrite (for example, converting a table model to incremental where appropriate).
  • DRY - repeated logic that should be factored into shared models or macros.
  • Testing & documentation - models missing tests or documentation.

How it works​

Human review​

Optimize is review-first - nothing changes your repo until you accept a fix, and fixes land as pull requests. See the Production safety model.