smoky sky
Ever asked an AI agent to build a data pipelines and watched it hallucinate column names, miss lineage, or generate SQL that doesn’t match your schema?
We built Altimate Code (open source, MIT) to fix that. It plugs into your warehouse + dbt project (models, lineage, schemas) so the agent works with real context and can actually execute dbt tasks end-to-end instead of guessing.
It’s been working well for things like model creation/refactoring, debugging runs, and navigating large projects. We’re seeing ~74.4% on ADE-bench (vs ~65% Cortex Code, ~40% Claude Code), but more interesting is just how it behaves on real projects.
Works with any warehouse + open AI API / Codex!!
Try it: npm install -g altimate-code
Examples: https://docs.altimate.sh/examples/