Datavor is an AI-native data pipeline tool built on a simple conviction: moving and understanding your data shouldn't mean handing it to someone else's cloud.
Five principles we don't compromise on.
Datavor runs on your machine. Your databases, your credentials, your pipelines — they never leave. There's no Datavor cloud your data passes through, no shared infrastructure, no egress fees. The Free tier makes zero outbound calls. You can verify it with a packet sniffer.
Everything Datavor learns about your data lives in a single SQLite file you own — ~/.datavor/context.db. No proprietary format, no lock-in. Open it with any SQLite browser. Copy it, inspect it, delete it. If you stop using Datavor tomorrow, nothing of yours is trapped anywhere.
No DSL to learn, no YAML to write, no drag-and-drop canvas to wrestle. You describe what you want in plain language, and your AI tool drives Datavor's tools to make it happen. The best interface for data work is the one you already use to think — words.
We tell you exactly what's stored and what isn't, what a feature can and can't do, and how things fail — not just how they work. Real-time CDC is hard; we document the failure modes instead of pretending they don't exist. Trust is earned by being straight with people.
Datavor is supported by the people who use it, not by investors who need an exit. That keeps the incentives clean: we build what makes the tool better for developers, not what maximizes a funding round. Free stays genuinely free. Pro is a fair price for a commercial license, not a hostage situation.
Enterprise ETL tools — Fivetran, Airbyte, the rest — are built for enterprises. They assume a cloud account, a billing relationship priced by row volume, a dedicated data team, and a willingness to route your data through someone else's infrastructure. For a large company, that trade is often worth it.
For an individual developer or a small team, it's absurd. You have a Postgres database and a warehouse and you want to move data between them, transform it a little, maybe replicate it in real time. You shouldn't need a SaaS contract and a per-row meter for that. You shouldn't need to learn a new orchestration framework. And you definitely shouldn't have to send your production data through a third party to do it.
Datavor started as a question: what would a data pipeline tool look like if it were built AI-native from day one, ran entirely on your own machine, and got smarter the more you used it? The answer became an MCP server with 47 tools, five database connectors, real-time CDC, a dependency-aware scheduler, and a persistent local memory — the Context Engine — that turns your AI assistant into something that actually knows your data.
It installs with npx datavor. There's no account to create. That's the point.
A small independent team that cares more about building a tool developers trust than about being a household name. We use Datavor ourselves, every day, on our own databases. The bugs we hit are the bugs we fix.
Questions, bug reports, feature ideas, or just want to say hi — we read everything.