Get it from the Microsoft StoreReads your files in place
Select a file or a whole folder and work through it. PDF and Word documents, including scanned, image-only PDFs up to 50 MB or 1,000 pages. No upload step.
Extract data from 500 documents. Without uploading one.
Lexapient runs on your desktop and calls Google Gemini directly. Your documents never pass through our servers, because we don't have any in the path.
01 · Direct to Google Gemini
02 · We never see your data
03 · Rerun single answer or row
04 · Reads files in place
Capabilities
Built for the way legal teams actually work
Three core capabilities, one workspace. More coming.
Bulk extraction with Playbooks
Define extraction rules once, then point them at 5 documents or 500. The rules do not change. Every extracted value comes back with a citation to the exact passage it came from, so any answer can be checked against the document.
Review that fits how you check work
Flag anything that needs a second look, mark a row reviewed as you clear it, and rerun a single answer or a single document when one needs another pass. A questionable answer costs you one cell, not the whole batch.
Zero-access architecture
We can't see your data. Not by policy. Not by promise. By architecture.
Direct to Google
Every API call goes straight from your machine to Google Gemini. No proxy. No intermediary. No man in the middle.
Local-only storage
Documents, Playbooks, prompts and results never leave your machine. Sessions are held in a SQLite file on your own disk. There is no copy on ours, because there is no ours.
Google doesn't train on it either
Runs go to Google's paid Gemini API, whose terms state that prompts and responses are not used to train Google's models and are not retained once the request completes. Your documents are nobody's training data.
Two endpoints. That's it.
Google's API, which receives the documents, and our service, which issues the key for a run and records what it cost. That record is a token count, never a file name, a prompt or a result. Verify it with any network monitor.
Process
Three steps. No upload.
01
Point at your documents
Pick a file or a whole folder from your machine. PDF and Word files are read directly. Nothing leaves the machine until you start a run.
02
Choose a Playbook
Pick a pre-built template or define your own extraction rules. Each rule specifies what to extract and the answer format: short answer, multiple choice, exact quote, or open answer.
03
Review and export
Documents process in parallel. Results land in a grid you work in. Double-click a cell to open its citation and read the passage the answer came from, flag anything that needs a second look, mark a row reviewed, or rerun a single answer without touching the rest of the batch. Export to Excel when it's signed off.
Every other tool in this category asks you to upload your files to their cloud, process them on their servers, and trust that nobody looks. That is a promise.
We built Lexapient so there is nothing to promise.
Pricing
Three credits free. Then pay for what you use.
There is no subscription, no seat count and no annual commitment. You buy credits in one-time packs and spend them only when you run an extraction; the longer the document, the more it costs, and the app shows you the figure when the run finishes. An idle month costs nothing.
How it is charged
$1 a credit
Credits are bought in one-time packs and cost less the larger the pack, down to $0.83. You spend them only when you extract, so there is no plan to pick, no renewal date and no minimum to hit.
- ✓ 3 free credits on every new account
- ✓ Buy credits in one-time packs, never on subscription
- ✓ Credits you have bought do not expire
- ✓ No monthly minimums and no per-seat fees
- ✓ Credit cost and balance shown after each run
A measured batch
0.75 credits
Twenty-five commercial lease agreements, twenty-one fields pulled from each: 525 extracted values, every one carrying its source citation. 0.75 credits, seventy-five cents, is the metered cost of the whole run, about three cents a lease.
- Your three free credits cover roughly a hundred leases of this length
- A $25 pack of credits covers several hundred more
- Shorter documents cost less: a single NDA runs about two cents
Questions
The six we're asked most.
Can you read our documents?
No. The application calls Google's API directly from the workstation. There is no Lexapient server between you and Google, so there is no point at which we could read a document, a prompt or a result.
How do we verify that claim?
Run any network monitor or check your firewall log while the app is working. You will see two destinations: Google's API, which receives the documents, and our management service, which issues the key and takes back a token count for billing, no file names, no prompts, no results.
Whose Gemini account is used?
Ours, so there is no Google account for you to set up. Each batch runs on a credential we provision for that batch alone: restricted to the text API, held in memory only, never written to disk, and deleted when the run ends. Bringing your own key is on the roadmap; tell us if you need it.
What happens to results when we close the app?
They are saved to a local database file on your machine, so the session is still there when you reopen it. Nothing is copied anywhere else. Export to Excel when the review is signed off.
Which file types are supported?
PDF and Word (.docx and .doc). Scanned, image-only PDFs are read as well, up to 50 MB or 1,000 pages per file; anything larger is skipped rather than partially read.
How do we get started?
Install it free from the Microsoft Store, point it at a folder and run your first batch. Your account opens with 3 free credits, so you can try it on real documents before spending anything.
Available now
Start extracting this afternoon.
Lexapient is live on the Microsoft Store. Install it, point it at a folder of documents and run your first batch. Your account opens with 3 free credits, so you can try it on real work before spending anything.