Start here

Getting started

From empty workspace to cited answers in about five minutes. No credit card, sample data included.

1 · Create a workspace

Sign upRegister with your work email and name your organization. Every organization is a fully isolated tenant. Files, embeddings, tables, and chats are scoped to it on every request.
Invite your teamAdd teammates from Settings → Members. Admins manage billing, connectors, and members; members use everything else.
Explore the sample dataNew workspaces can load a demo dataset (sales, invoices, telemetry, policies) so you can try real questions before connecting anything.

2 · Add your data

Upload files

Go to Files and drop in documents: PDF, Word, Excel, CSV, Markdown, text, and more. Organize them with folders. Two things happen on upload:

Text is indexed. Documents are chunked and embedded so chat can retrieve and cite exact passages.
Tables become tables. Spreadsheets and tabular PDFs are ingested as real, typed SQL tables. This is what makes numeric answers exact.

Connect your tools

Go to Connectors and connect Gmail, Google Calendar, Google Drive, Notion, Slack, Confluence, Salesforce, HubSpot, or a database (PostgreSQL, MySQL, MongoDB). Syncs return instantly and run in the background; only new or changed content is ingested on re-sync. See Connectors for per-tool setup.

3 · Ask your first question

Open Chat and ask in plain language. Good first questions:

What was our total revenue by region in 2025?
Which invoices over $10k are unpaid past 30 days?
What's our parental leave policy in Germany?
How do I call the invoices API? Give me a curl example.

Watch the reasoning steps stream in, then check the answer's receipts: numbered citations open the exact source passage, and the SQL chip on computed figures opens the query that produced them.

Tip Flip to Visual (top of the chat) to watch retrieval, routing, and reasoning animate in a full-screen view. It's the fastest way to understand how answers are built, and a great demo.

4 · Go further

Build an agent. In Studio, describe what you want automated, like "classify inbound emails and extract the order number", and the copilot builds the agent and its flow.
Generate a dashboard. Describe the dashboard you want over your data; revise it in plain language. Every revision is audited.
Plug in your own tools. Connect an MCP server, trigger flows by webhook, or let agents call out; see MCP & integrations.