Documentation
Your company's brain, on demand.
Vectropic.ai turns your files, tools, and databases into one assistant that answers with citations, exact numbers, and expert-level reasoning. This documentation covers the whole product, from your first question to building custom agents.
Getting started →Create a workspace, add data, ask your first question. About five minutes.
Chat & answers →Citations, SQL-computed numbers, expert agents, Visual mode, and voice.
Connectors →Gmail, Calendar, Drive, Notion, Slack, CRMs, and direct databases.
Studio →Build custom agents and multi-step flows; run history and audit for everything.
What Vectropic is
Most AI assistants predict an answer and hope it's right. Vectropic retrieves, computes, and reasons, then shows its work:
- Answers with receipts. Every reply cites its source. Click a citation to open the exact passage, highlighted, or the exact spreadsheet rows.
- Exact numbers. Spreadsheets and PDF tables become real SQL tables. Quantitative questions are answered by running a guarded, read-only query you can inspect, never by predicting digits.
- Experts that reason. A router classifies every question by domain and difficulty. A specialist (financial analyst, technical writer, legal, HR, sales, or data) answers in the right voice. Hard questions are decomposed, researched in parallel, and synthesized.
- Honest gaps. When your data doesn't cover a question, the answer says so and points to what it does cover.
The pieces
| Surface | What it does |
|---|---|
| Chat | Ask anything in plain language. Answers stream in with live reasoning steps, citations, and the SQL behind any computed figure. |
| Visual | A second, full-screen view of the same conversation that animates retrieval, routing, and reasoning as they happen. Great for understanding, and for demos. |
| Files | Upload PDFs, Word, Excel, CSV and more. Organize with folders; tabular files become queryable tables automatically. |
| Connectors | Background, incremental sync from Gmail, Google Calendar, Drive, Notion, Slack, Confluence, Salesforce, HubSpot, and direct Postgres / MySQL / MongoDB connections. |
| Studio | Describe an agent in plain English and get a working one (persona, tools, and a multi-step flow) executed on a durable queue with full run history and audit. |
| Dashboards | Generated, revisable dashboards over your data, with their own audit history. |
How answering works
Every question moves through a three-layer agent pipeline:
- RouteOne classification decides the domain (financial, technical, legal, HR, sales, data, general), whether the question needs documents or SQL, and whether it's simple or multi-step.
- Retrieve & computeRelevant passages are retrieved from the vector index and graded; quantitative questions compile to guarded, read-only SQL over your ingested tables. Weak evidence is re-routed to a better source, not papered over.
- Reason & composeComplex questions are split into sub-questions, researched in parallel, and synthesized. The matching domain expert composes the final answer, with citations and derivations shown for anything computed.
Each of these steps streams into the UI while it runs, so "thinking" is never a spinner. It's the actual plan and progress.
Where to go next
- New workspace? Start with Getting started.
- Wiring up your tools? See Connectors for per-tool setup (including Google OAuth and Notion sharing).
- Automating work? Studio covers agents, flows, run history, and audit.
- Bringing your own tools? MCP & integrations covers MCP servers, webhooks, and outbound actions.