Ask your own code and data a question in plain language. Get a cited answer you can trust — without waiting on an engineer.
01 A query layer over your own systems — that isn't a person
In a data-rich organisation, the knowledge of how a system works and what a metric actually means lives in the codebase, the warehouse models, and the heads of a few engineers. Every time ops, program or exec staff need that knowledge, it routes through engineering as an interrupt — a ticket, a Slack ping, a meeting. Engineering becomes a human query layer over the company’s own code and data.
Lodestar absorbs that load. It is a governed, sandboxed assistant that reads your code and queries your data platform, then answers questions about both in plain language — with the sources cited, so the answer is trustworthy enough to act on.
It is the production-grade successor to an assistant we built and proved with our client. What we learned there — about governance, citation and getting an answer someone can actually trust — is what Lodestar is built from.
02 Governed, cited, and built for the people who aren't engineers
Code and data, together
Not just a code assistant and not just a BI tool — Lodestar reads the codebase and queries the warehouse, so it can answer 'how does this work' and 'what does this number mean' in the same breath.
Cited, or it doesn't count
Every answer shows where it came from. A figure you can trace back to its source is a figure you can take to a decision; one you can't is just a guess with confidence.
Governed by default
Read-only, sandboxed, and fenced by your own access rules. Confidential sources stay confidential, and the assistant tells people plainly when something is out of reach rather than guessing around it.