Start with what you're trying to find out.
You don't have to learn a system to get an answer out of one. These pages say what Overscore can do and when each part is the right one. Describe what you need in plain words; the product, or your agent, picks the right door. And every question you ask, every correction you make and every dashboard you build teaches it, so the second month is better than the first.
Ask a question
Plain words in. You get the number, the table it came from, and the SQL folded underneath if you want it. The answer reads what it knows about your business before it reads the data.
Connect a source
Shopify, Stripe and QuickBooks take a pasted key. Your own Postgres links live. Spreadsheets and files drop in. Everything syncs on its own and tells you when something's off, with the fix ready.
The inbox
What broke, what drifted, and what it heard you say that needs a yes. Each card is one sentence with a reply box, and your reply becomes a fact it keeps.
Worklists
Any query with marks becomes a list people can work through and check off, with a note on each row. The marks live with the data, so the next report knows.
Dashboards
Built by you or your agent, in your company's theme, on a phone. Every number on it carries the definition it used.
Context and definitions
What you say becomes a fact it keeps. A number computed in two places becomes one definition with an owner. When it gets something wrong and you correct it, that correction is kept too. Every answer, every dashboard and every agent reads these first.
For whoever's technical
Nothing above needs this section. It's here for the person on your team, or the consultant, who wants to see under the hood.
Tables and SQL. Every table, where it came from, and what people know about it, in one tree. A query editor with history and saved queries. Ask, and it writes the query.
Models. SQL that runs on a schedule and lands as a table, with tests. Export the lot as a dbt project any time.
Agents and keys. One address for Claude, Cursor or any agent. A key fits one dataset at a time, at add-only, read, or read-write, and can mask names and emails. Everything an agent does is logged like a colleague's work.
The Postgres door. On Paid, the warehouse answers as a Postgres database, so Metabase, Power BI and Tableau connect the normal way.
The CLI and the API. ovs on your laptop does what the app does. The HTTP API does the same for scripts. Reference pages are coming; until then ovs --help is complete.