Pipelines and schedules that run themselves.
Schedule a query, export it to a bucket, or run your DBT project from Git — on the same engine as everything else, with alerts when a run fails or data goes stale. Your agents can set it all up for you.
Scheduled queries and exports, in your own timezone.
Register any SQL on a cron schedule — a daily rollup, an hourly incremental load, a nightly export of a report to S3, GCS, Azure or R2 as CSV, Parquet or Excel. Each run is sized like any other query, shows up in query history, and alerts you if it fails.
- Cron plus IANA timezone — runs follow your local time, not UTC
- COPY a result straight to your own bucket on every run
- Pause, resume, hand a job to a teammate, or replay any run in Studio
DBT pipelines straight from your Git repo.
Connect a repo and its thunderduck.yml decides what runs when. Production runs whatever is merged to the default branch; a branch run builds into dev catalogs so you can check the tables before you open the pull request. Every model's freshness is tracked against its schedule.
- Run, test, seed and snapshot on the same engine — no separate warehouse
- Branch runs write to dev catalogs, never to production
- Freshness per model, and an alert when one goes stale
Hand the busywork to your agents.
Agents schedule it
Ask Claude or ChatGPT to run a query every morning and export it to your bucket. Over MCP it checks the SQL, prices it and creates the job — owned by you.
Agents build models
An agent pushes a branch, runs it against dev catalogs, checks the tables and opens the pull request. Production only changes when you merge.
AI explains failures
When a run fails, Claude reads the query, its plan and the tables, says what went wrong and who can fix it, and suggests the corrected SQL.
Guardrails on every run
Runs use the owner's access, never more. Each one is cost-checked, capped by the company's timeout, and counted against its concurrency limit.
