Open Source.
Paid.
Data Engineering Agent.
Same catalog across every tier. Start with Lite — the whole platform, Apache 2.0. Upgrade for enterprise access controls, or bring the agent that owns the pipeline end-to-end.
- Full catalog — operators, actions, connections, payloads, skills, usecases
- AI operator that generates, runs, and repairs pipelines
- MCP server (~25 tools) for Claude Code and any MCP client
- `inspect_data` across Postgres, Athena, Redshift, BigQuery, S3, GCS
- Multi-user — bring your whole team, no seat cap
- Community marketplace (import & publish)
- Docker Compose install, Blue-Green production deploy
- SSO / SAML
For teams that need item-level permissions, groups + SSO/SAML, a private self-hosted marketplace, admin-side MCP restrictions, or managed deployment in your own VPC (data + compute stay in your account).
Priced per engagement — this is the tier that funds the research today. Every conversation gets a real answer.
- Real tools, not just chat — runs, inspects, and fixes forward like an engineer
- Composable skills — team, infrastructure, pipeline, data contract
- Learns from corrections, humans approve — no drift without a sign-off
- Runs on your existing LeastAction catalog and operators
- Works against Postgres, dbt, Airflow, and any connection in the catalog
LeastAction is a research effortat heart. The paid tier is what pays for the team to keep working on it. If the open-source + community model becomes self-sustaining, the enterprise features move to Apache 2.0 — not a promise for a specific date, just the direction we’re heading. We’d rather have the conversation about that honestly than pretend the paid tier is a permanent institution.
Run a million tasks or one — same price.
Bring your own Anthropic, OpenAI, or Gemini keys.
Import and publish freely on every tier.
Self-host means your data never leaves.
Frequently asked.
Honestly answered.
Can't find what you're looking for? Drop us a line.
Not sure which fits?
We'll help you decide.
Twenty minutes with our team gets you a clear answer — and a working pipeline by the end of the call.