Product
Document Processor
An upload widget that reads a document and hands back structured data — a summary, key points, extracted names/dates/amounts, and a guessed document type — instead of a wall of raw text.
Included in every plan, configure it yourself. Want it done for you instead? See this as a service.
How it works
- 1
Someone uploads a file
PDF, TXT, DOC, or DOCX, up to 10MB. Text is extracted server-side and the file is deleted immediately after — nothing lingers on disk.
- 2
Your AI key extracts structure
The extracted text (truncated to a safe length to control cost) is sent to your chosen AI provider with instructions to return summary, key points, entities, and document type as structured JSON.
- 3
You get structured output
The result is stored and shown in your dashboard, and can be customized with your own extraction instructions instead of the default.
Who it's for
- Law firms triaging incoming contracts or filings
- HR teams processing applications or onboarding paperwork
- Finance teams pulling numbers out of invoices
- Anyone drowning in PDFs who wants the highlights, not the whole document
What it doesn't do
Long documents are truncated before the AI call (about 8,000 characters) to control cost — it summarizes what it sees, not necessarily the whole file if it's very long.
DOC/DOCX support is best-effort text extraction today, not full formatting-aware parsing — for guaranteed fidelity, PDF or TXT are more reliable.
It extracts what's asked for by the prompt — the default categories are summary, key points, entities, and action items. Anything outside that structure needs a custom prompt.
It doesn't verify the accuracy of extracted numbers or dates against a second source — always spot-check anything used for a financial or legal decision.