Capture
Send a PDF or image of a receipt from WhatsApp, or upload it here.
Capture receipts directly from WhatsApp, let the agent extract merchant, totals, and line items, then ask spend questions in plain language — all from one workspace.
Four stages, fully automated end to end.
Send a PDF or image of a receipt from WhatsApp, or upload it here.
The agent extracts merchant, totals, taxes, and individual line items.
Original file goes to Supabase Storage; structured data lands in Postgres with embeddings.
Ask spend questions in plain English — answered with SQL plus semantic search.
Live state for the active user.
Pick a quick action or jump to chat.
Send messages or upload receipts as a WhatsApp user.
Receipts are persisted in Supabase Storage and indexed in Postgres + pgvector. Use chat to add new ones or run searches.
Review pending uploads, saved receipts, and generated invoices for the linked user.
No receipt uploads or invoices for this user yet.
Outgoing invoices created from WhatsApp or the website are saved under the active user and available here.
No generated invoices yet. Ask the chat to create one for a client.
Processing stages that handled the most recent message. Click a step for logs.
Classifier output for the last user turn.
Last LLM call.
Identity used for the current conversation.
Your signed-in phone number is used as the WhatsApp account for uploads, analytics, and chat.
To change this number, update and verify the phone number on your sign-in account.
Manage pgvector embeddings for semantic search over items and invoices.
Postgres health and user-scoped counts.