Lexabrief
Documentation

Getting started

Six steps take you from an empty account to an answer written out of your own documents. Nothing below costs money until step six.

1. Create an account

Sign up at /signup. Continue with Google, or use your email address and create a password. Those are the only details account creation asks for. If you sign up with an email address, a verification link is sent to it and the workspace stays closed until you follow that link. The screen has a Resend button and re-checks itself when you come back.

New accounts receive 100 credits, once. Credits never expire, and they are spent only when a model does work for you: uploading and reading documents, OCR on scans, and storage inside your plan are free. Free accounts get 500 MB of storage.

2. Open a case

The All cases screen. One row per matter, each showing the case name, whether a docket number is on file, the court, a count of documents and conversations, the next deadline, and when it was last touched. A filter box and a New case button sit above the list.
Your workspace: every matter you own or have been shared on, with what is in it and what is next.

Your first sign-in provisions one private matter called "Starter Case" so there is somewhere real to try this. For actual work, use New case on the home screen: it takes a title and a short description, and the practice templates set the matter type for you.

The New case dialog open over the matter list. It offers four practice templates (employment wage and hour, personal injury, business and contract, unlawful detainer), a Caption field for the case name, and an optional Notes field described as a few lines of context for Lexa.
Opening a matter. The caption is the only thing it needs; the notes are read by Lexa as context.

A case is the unit of everything. Documents, conversations, drafts and citation checks all live inside one case, and Lexa never reaches into another one. Inside a case there are three places to be: Chat (ask lexa, and draft with her), Documents (everything filed in this case), Drafts (what lexa has written in this case), Case info (parties, court, and key dates).

3. Upload the record

The Add documents dialog open over a case's Documents view. A large drop zone reads "Drop files here, or click to upload" and lists the accepted kinds underneath: PDF, DOCX, TXT, MD and image. A line below the zone says the files are private to this case and that Lexabrief does not train on them.
Upload documents is in the case sidebar in every view. Drag a batch in, or click to pick.

Upload documents sits in the case sidebar in every view, and the empty documents table takes a drag and drop. Upload the pleadings, the discovery responses, the deposition transcripts, the exhibits, the order you are responding to. This is the record Lexa answers out of, so what you leave out, it cannot see.

  • JPEG image
  • PNG image
  • GIF image
  • WebP image
  • TIFF image
  • PDF (including scans, which are read with OCR)
  • Plain text (.txt)
  • Markdown (.md)
  • CSV (.csv)
  • Word (.docx)
  • Excel (.xlsx)
  • Email (.eml)
  • Outlook message (.msg)

Up to 30 files per batch, 200 MB per file. Scanned PDFs are read with OCR, which takes longer than a text PDF of the same length.

4. Wait for processing

A case's Documents view once the record is in. The header reads "6 documents, All ready". Each row carries the document's own title, the filename it arrived as, a one-paragraph summary Lexa extracted from it, the docket number and date where the paper has them, and a green Processed status. A folder tree runs down the left.
The header states the count as it changes, and every row says where that document got to. Ready means the text was extracted and indexed.

Each uploaded file goes through two stages before Lexa can retrieve from it: the text is extracted (with OCR where the page is an image), then the text is indexed for search. A document is ready only when both have finished, and the header on Chat and Documents states the count as it changes, for example "4 of 6 documents ready".

How long depends on page count and whether the pages are scans: a short text PDF is usually a minute or two, a long scanned transcript can run to several. Processing continues if you close the tab.

You are not blocked while this happens. The message box stays live, so a question about the law is answerable while exhibits are still indexing. A question about the record is worth waiting for.

5. Find out what to ask

The capability picker, opened from the case sidebar and anchored just above the message box. It says "Tap a question to load it into your message. Edit it, then Send." Under it are three groups of questions in a lawyer's own words: understanding what is on file, researching the law, and writing then checking a draft.
Everything the product does, as questions rather than feature names. Tapping one loads it into your message box; it does not send it.

What can Lexa do? in the case sidebar opens the full catalog of what the product does, grouped into "Understand what's on file", "Research the law", "Write it, then check it". Tap a question to load it into your message. Edit it, then Send.

That last part matters: tapping a question never starts a run. The sentence lands in your message box, you edit it into your matter's own facts, and nothing is spent until you press Send.

6. Ask your first question

On a case with documents in it, the question that teaches you the most about what you have is this one:

Example prompt

Read what's on file and tell me what this case is really about.

Reads the whole record and gives you the shape of the matter, with the pages behind it.
The message box at the foot of a case conversation, holding the sentence "Read what's on file and tell me what this case is really about." Under it sit an Attach or ask control on the left, the model picker showing Flash, and the claret Send button on the right.
The sentence sits in your box until you press Send. Edit it into your own facts first; nothing is spent before the press.

What to expect while it runs: Lexa works in steps and narrates them, so you will see it reading documents, searching, and composing rather than a blank spinner. A short question about one document comes back in a moment. A research or drafting run works for minutes, sometimes many, and keeps running if you switch views or close the tab.

The model picker beside the message box offers DeepSeek V4.1 Flash (fast, capable work with low relative usage) and GPT-5.6 Luna (efficient reasoning with a large working context) and GLM 5.3 Flash (fast agentic work with low relative usage) and Gemini 3.8 Flash (strong reasoning for complex research and drafting) and Claude Sonnet 5 (deep analysis for the most demanding work). DeepSeek V4.1 Flash is the default and is the right choice for most questions; the heavier model reasons over more and costs more per answer.

When Lexa writes something long, it opens a document beside the chat instead of pouring prose into the transcript. That document is editable, it is where the citation check runs, and it is what exports to Word.

A case workspace with the conversation on the left and a finished memorandum open on the right as a paper page. The memo carries a caption, a Question Presented and a Short Answer; two of its citations are tinted amber where the check has something to say about them. The panel header offers Draft and Citations tabs, and the Citations tab carries a count of four.
The conversation on the left, the document it wrote on the right. The tinted citations are the check's own marks, and each one jumps to its entry in the review. From here, read the workflow.