Quickstart

1. 1.0Create an account
   Signing up gives you 2,000 credits, which is 2,000 pages of reading — enough to put a real set of documents through and see the answers. No card, and there is no subscription to cancel. The form asks one question about your work: which kinds of documents you deal with. Tick as many as apply.
2. 2.0Choose what to start with
   That answer becomes a short list of starter tables — one per kind of document you named, already ticked. Continue and you land in a workspace with those tables built, columns and all. Skip it and you get an empty account instead, which is the better choice if you would rather look around first and make the workspace yourself.
3. 3.0Upload your documents
   Drop the files in. Each one is read, split into pages and indexed, at one credit per page. That happens once: every table you build later asks questions of documents that are already read.
4. 4.0Read the columns before you run them
   Every document in the workspace becomes a row. If you took a starter table its columns are already written — read them first, because they are illustrative prompts meant to be edited to fit your documents rather than definitions to adopt. If you started empty, create a table now: it opens with no columns and offers the same catalogue.
5. 5.0Write a column
   A column is a question asked of every row, plus an output type. Ask for one thing you actually need — "What is the notice period for termination for convenience, in days?" — and set the type to number, because that is what you will want to sort by.
6. 6.0Run it
   Select the rows and columns you want and run those. Cells resolve independently, so answers arrive as they land rather than all at once. Start with two rows: it costs almost nothing and tells you whether the column is asking the right question.
7. 7.0Check an answer
   Click a cell. The citations panel opens the page the answer was read from, so you are checking one page rather than re-reading the document. If it is wrong, type over it — the original is kept underneath.
8. 8.0Run the rest, then export
   Once a column reads correctly on two rows, run it on the remainder. Export the finished table to XLSX or CSV when you need it somewhere else.

Two things worth knowing before you start

Tables start in auto mode, including the ones you were given. A table in auto mode extracts each new document added to its workspace without anyone pressing Run — which is what you want for documents that keep arriving, and not what you want while you are still deciding whether a column asks the right question. It matters more at step 3 than it used to: if you took three starter tables, your first upload runs three tables against every file in it, and that is a real amount of credit spent before you have read a single answer. Switch them to manual in each table’s settings until the columns are settled. See auto mode.

A new account is limited to 100 pages per document. That limit rises with lifetime spend rather than with a plan, and a document over it will not ingest. If your first upload is a 400-page credit agreement, read tiers and limits first.

Where to go next

Core concepts is the shortest way to understand why the loop is shaped like this. Columns is the page to read properly, because a disappointing table is almost always a column problem rather than a document problem. And the use cases are seven worked setups you can copy outright.

Start free