Follow one article down the line.
Libril isn't a prompt box with a subscription attached. It's an assembly line: ten stages of work between your two-sentence brief and a published, illustrated, cited page. Scroll and watch it run — the panel on the left tracks where you are, how long it has taken, and what it has cost so far.
Brief
A topic and a line about what you want is genuinely enough — your project already holds the company overview, the tone of voice, and the document library, so Libril starts with context a freelancer would need a week to absorb.
When the article matters, take control: angle, target reader, primary and secondary keywords, internal links with anchor guidance, and hard constraints on what the piece must never claim.
- Simple mode — a topic and an objective
- Detailed mode — keywords, audience, reading level, intent
- Internal links you want placed, with anchor guidance
- Claims to avoid and competitors to name or not name
- Bulk briefs imported from Excel or CSV
- Topic
- Choosing a heat exchanger for cold-climate retrofits
- Objective
- Help specifiers pick between plate and shell-and-tube below freezing
- Primary keyword
- cold climate heat exchanger
- Internal links
- /hx400 · /retrofit-guide
- Knowledge base
- 412 documents attached
Research, five passes
Nothing is written until the research is done. Five passes run in sequence, each with a different job: establish the facts, mine authoritative sources, imagine what a real reader wants to know, answer those questions from real pages, then find what every competing article is missing.
In parallel, Libril reads your own library. Point a project at a folder and it indexes what is there — fourteen formats, on your machine, no import step — then pulls the documents relevant to this article and prefers them over anything it found on the open web. Your spec sheet beats a competitor's blog post, every time.
- Fourteen formats read natively — PDFs, Word, spreadsheets, Markdown, HTML
- Read, indexed, and matched entirely on your own machine
- Auto-organized by whatever dimension matters to your business — a state, a product, a service line — with one-click corrections
- Drop a new file in the folder and it is picked up live. No restart, no re-scan
- No source, no claim: if research didn't find it, it doesn't go in the article
Outline
The outline is built around topic coverage and entity relationships, not a keyword sprinkled through eight headings. Sections earn their place by covering something the reader actually needs, ordered the way people read: answer the question they arrived with, then earn the depth.
External links get a plan rather than a sprinkle. Each candidate is chosen for reader value and source authority — government, educational, and standards bodies rank highest — and each one carries a documented reason it earned its place, woven into the outline before a word is written.
Draft
Writing happens here — after eight other stages, not instead of them. The draft follows the outline, uses the verified research, and speaks in the tone of voice stored on the project, so it reads like an extension of your expertise rather than a generic AI voice you then have to rewrite.
Facts arrive with their citation attached at the moment they are used, not reconstructed afterwards. That is why there is nothing to fact-check.
Self-edit
Libril edits its own work before showing it to you. Redundancy comes out, hedges get committed to, structure tightens, and anything the outline promised but the draft skipped gets written.
This is the stage most tools skip, and it is the reason the draft that reaches you needs fifteen minutes rather than an hour.
Humanize
The humanizer came out of months of comparing hundreds of human and machine-written documents to isolate the repeatable tells — not a vague sense that something reads wrong, but the specific habits that make a reader stop trusting a page.
We are not trying to fake a human or beat a detector. We are removing the tendencies people now recognise as AI slop. Passing detection turns out to be a side effect of writing that is actually good.
Illustrate
Libril illustrates the article it just wrote, from that article's own content rather than a generic stock prompt. Each project carries a visual style guide built by looking at screenshots of your own website, so colours and treatment follow your brand automatically.
Every visual lands at the heading it belongs to. Two variants arrive for each one, and any single image can be regenerated from its saved prompt.
Publish
Finished articles go straight to your site — one at a time, or a whole batch from the publish dialog, images included. You choose which assets travel with each article.
Clean, semantic HTML export remains the universal fallback: headings are headings, lists are lists, citations are real links, and nothing is locked in a format that needs Libril to open it.
Account
Every call's real spend is tracked per article and split by generation mode, then surfaced in your project dashboard next to word counts and publish status.
Most tools give you a credit balance and an estimate. We don't estimate your cost per article. We show it to you.
| Research · five passes | $0.62 |
| Outline & link plan | $0.21 |
| Draft | $0.29 |
| Self-edit | $0.16 |
| Humanize | $0.14 |
| Visuals | $0.08 |
| Total, this article | $1.50 |
Express when you need it now. Economy when you're scaling.
Same nine stations, same output. The difference is whether you are watching.
Express
Real-time generation, article by article. You see each station complete and a finished draft lands in ten to twelve minutes. This is the mode for the piece you need this afternoon.
Economy
Batched generation at roughly half the API cost. Start a batch, close the app, come back to finished drafts — an Economy run survives a restart mid-flight and picks up where it left off. This is the mode for a content calendar.
Fifty briefs, one paste
Batch briefs import straight from Excel or CSV, so a quarter of planned content becomes a queue in a single step. Import fifty briefs on Friday in Economy mode and collect fifty drafts at half price, each one with its own research, visuals, and per-article cost.
- Choose the mode per batch, not per licence
- Per-station progress for every article in the queue
- Publish the whole batch when you're happy with it
No server. No seat. No expiry.
The line runs on your desk. Here is exactly what that means.
- Licence
- $299, once. You install it and you keep it. There is no server to switch off, no seat to reprice, and no tier that quietly removes a feature you were relying on. If we ever shut down, your copy keeps working and we have committed to open-sourcing the code.
- Model access
- Bring your own key. You connect your own provider account and pay published rates for the intelligence you use. We never sit between you and the model, so we never need to mark up tokens, ration them, or bundle them into credits that expire.
- Your data
- Local, by architecture. Projects, briefs, your document library, research, drafts, brand voice profiles, and finished articles live in a database on your machine. Reading and indexing happen on-device. There is no cloud copy to breach, subpoena, or train on.
- Cost basis
- Wholesale, and measured. About $1.50 an article in Express and half that in Economy — the real figure, tracked per article and shown to you, rather than an estimate against a credit balance.
- Scale
- No quota, ever. Unlimited projects, articles, and brand voice profiles on one licence. No volume pricing, no overage, no upgrade conversation, because there is nothing to meter.
- Portability
- HTML that outlives us. Direct publishing to WordPress, Shopify, and Wix, plus clean semantic HTML export for anything else. Nothing is trapped in a format that needs Libril to open it.
- Updates
- Free forever, and delivered. The knowledge base, Economy mode, spreadsheet import, image generation, direct publishing, and cost tracking all shipped after launch and reached existing licence holders at no extra cost.
- Requirements
- macOS 12+ or Windows 10+. 64-bit, 4 GB of RAM, and about 1 GB of disk plus whatever your own content needs. No browser extension, nothing running in the background.
Stop renting your
content engine.
Download Libril, connect your own API key, and write your first five articles today.