In the current broken system, an agent skims your query letter and passes while clearing out their inbox, or an assistant skims five pages and decides your book's future in ninety seconds. Four months later you get a form letter. It's slow. It's a relic. It's indefensible.
We're inverting the model. The old way, authors push manuscripts at agents and hope. Our way, agents pull from a filtered, ranked, pre-qualified pool.
Better for both sides of a market that currently works for neither. If your submission stands out, they'll know about it.
Simple, and deliberately weighted in your favor: we only keep money from authors we told "not yet" — and we hand them a credit to come back.
Anyone can paste a manuscript into an AI and publish whatever comes back. That's one opinion, from one model, with nothing to tell you how much to trust it. Here is what happens instead.
A first pass reads the whole book to work out what it actually is — the category, the sub-category, who it's for, and what it's attempting. Nothing is scored yet. This step exists because being measured against the wrong yardstick is the fastest way to get feedback that's no use to you.
That reading assembles an evaluation framework for your book: the craft standards its genre is actually held to, the published titles it will be measured against, and where the bar sits in that category. A mystery is judged on whether the clues were fair. A thriller is judged on the opposite. The framework is built per manuscript, not pulled off a shelf.
Several AI readers from different companies read the entire book against that framework, independently, without seeing each other's scores. They disagree — that's the point, and it's why we use rival models rather than one. A supervisor then takes the disagreement into a structured debate, weighs which argument is better supported by your text, and settles it in writing.
A score means nothing unless it's pegged to something real. So we run the system over public-domain books whose fate is already settled, from classics to novels that sold well and vanished, and check our ranking against theirs. The first time we did, it told us our model was wrong. We rebuilt it.
Agents scanning the marketplace know that everything in it has been pre-screened against a rigorous bar.
The bar is 85 out of 100 on quality, set at a professional standard. If you clear it, agents and publishers know it's ready.
Not clearing is useful too. You'll know where you stand, what's holding the book back, and what to fix first. That's more than a hundred queries will ever tell you.
The system gets sharper over time, and no model is ever trained on your book. What sharpens it is re-scoring books we've already scored, ours and yours, to catch the scoring drifting. That happens inside our own encrypted database. Your manuscript is never shown to a person, never used to teach an AI to write, and goes to nobody, no company and no agent, unless you say so.
And the numbers we publish about how well this works are on our methodology page.
One number, out of 100, judged against the standards of your genre. Commercial and IP potential are assessed separately.
Three scores: quality, commercial viability, and potential IP value. Plus the published books you were measured against, what's working, what's holding it back, and the strongest case for and against your manuscript.
If you didn't clear the bar, you get the five changes that would move your score most — named plainly, with the chapters they live in. Not "tighten the pacing." Which chapters drag, and what to do about them.
Quality comes back as a number out of 100, because it decides whether your book reaches agents and it's the one we hold to that standard. Commercial and IP value come back as a band — low, average, high, or extremely high. That's deliberate: nobody reads a book differently at 47 than at 51, and printing two digits would claim a precision those two judgments don't have.
A fair, fast way to find out where you actually stand — and if you're ready, to be discovered rather than to keep knocking on doors. No connections, no perfect query letter, no waiting four months to learn nothing. Your book is judged on the writing, by the same standard as everyone else's.
A marketplace of pre-screened work, every book read in full and scored — not skimmed. Browse by genre, ranked by quality, with commercial and adaptation potential shown on every listing. Each one comes with full coverage and scoring, so you can spend your reading time on books that already cleared a professional bar.
Not with agents, not with publishers, not with anyone — unless you approve it, one party at a time. Submitting transfers no rights. You keep everything you walked in with.
No model learns your voice, your sentences, or your story. Your book is only ever sent to providers barred from keeping it or training on it — a restriction we set on every single request, not a setting we trust to a default.
Your manuscript stays in encrypted storage so our scoring keeps getting sharper: checking judgments stay consistent, testing new versions against books we've already read. Access is restricted to that purpose and logged.
One click, permanently gone from our systems. Your score stays yours — it's built from numbers, not from your words.
He got a 79. Six points short. He agreed with every note.
The first manuscript our panel ever read belonged to the founder's husband. Two years of work, a novel he genuinely believed in — and the verdict came back: promising, not ready. The ideas are the best thing in it. But the middle drags, the plot leans on coincidence where it should turn on choices, and the ending arrives before it's been earned.
He read it and said: yes, all of that is true. He's on his sixth rewrite.
Then he did something we hadn't planned for. He sent us three drafts of the same book — an early one, and two written after he'd read a book about how a novel is supposed to be built. The early draft was rougher, and it scored higher on originality, on dialogue, and on voice. The draft he'd engineered scored better on structure, on pacing, and on its ending. He had sanded down the strangest parts to make the book more conventional, and the numbers showed him exactly what that had cost: he'd bought craft with the things that made the book his.
He'd been rewriting for two years and hadn't seen it — and there is no version of a rejection letter that would have told him. That's the part we didn't expect to be building. Not just whether a book is ready, but which of its faults are worth fixing and how much of its uniqueness is worth protecting.
Most submissions won't clear the bar. Ours didn't. That's what makes clearing it mean something.
We're calibrating. You get the full evaluation at no cost — pass or fail — and we get the data that makes the next thousand assessments sharper. When the hundred are gone, it's $49.99.
.docx, .pdf or .txt · 20,000–200,000 words · five-day turnaround
Your manuscript is encrypted at rest and never used to train anyone's model. Nothing is shown to an agent unless you choose it.
We're in beta and genuinely want to hear it — a broken page, a score that seems off, a report that didn't make sense. A person reads every one of these.