FireQuill is not a prompt box. It is a full long-form writing system — a living story bible, specialist editors that flag (never overwrite) your prose, and generation that remembers your characters, continuity, and voice from chapter one to the end.
Voice, psychology, knowledge, relationships, arc trajectory, all versioned per scene. Time-travel through your protagonist’s evolution.
Every character in your manuscript carries a structured object that the engine maintains alongside your prose. As you write, FireQuill’s extractor agent reads each scene and proposes updates to that object: a new fact the protagonist learns, a relationship that shifts, voice that slowly hardens. You accept or reject each proposal. The engine snapshots the state per scene so you can rewind.
For non-fiction, the same mechanism powers an Expert Panel: each domain expert in your book carries methodology, blind spots, key sources, and consultation voice. The editor agents draft as them, citing sources, surfacing perspective.
Stress-test dialogue with your protagonist. Ask your contrarian expert to demolish your central argument. Every voice in your book becomes a reachable collaborator.
Open the Cast tab in the editor and pick a character. They reply as them: voice samples plus psychology plus current arc state plus every recent extraction packed into the persona. Test a piece of dialogue before you commit it. Throw a hypothetical at your antagonist and watch the reaction. Workshop a confrontation scene from the inside.
Non-fiction’s Panel tab does the same with experts. Pick the contrarian on your panel and ask them to find the weakest claim in your latest section. Now your “kill your darlings” pass has muscle.
Developmental, line, plot, continuity, argument, evidence, counterargument, audience-fit: eight specialists trained as specialists, not as generalists pretending.
Most AI editors are one model trying to be everything. Ours are eight agents, each prompted, calibrated, and tested against fixtures for one job. The line editor doesn’t try to spot plot holes; the plot editor doesn’t critique sentence rhythm. Each returns a verdict, severity, and confidence: flag, don’t block. They never auto-correct your prose. They surface notes in a triage inbox; you accept, defer, or dismiss.
The system also catches itself: every dismissed flag is recorded; every override logged. Patterns become signal: characters whose voice flags get dismissed repeatedly get a “voice drift might be authorial intent” treatment.
Reads each chapter for pacing, cohesion, setup-payoff health, and act-structure fit.
Diction, rhythm, dead verbs, modifier sprawl, dialogue-tag overuse. Sentence-level craft.
Causal-chain checks across scenes. Setup, payoff, motivation visibility, agent of change.
Knowledge state, timeline, object location. Catches the things you’ll never notice yourself.
Per-chapter argumentative coherence. Claim, support, warrant, hidden assumption.
Claim-evidence proximity, source quality, citation-style consistency.
The objections your chapter doesn’t engage with. Surfaces fair-minded pushback.
Reading-level enforcement, jargon density, assumed-knowledge calibration.
Each verdict returns pass / warn / fail × cosmetic / notable / serious / breaking × low / medium / high confidence. Not “great chapter, consider tightening pacing.” Calibrated, scoped, traceable, false-positive-tested.
Plot archetype. Narrative framework. Premise. Themes. Cast or panel. Frame. Comparable works. The book’s bones, fully articulated, before you write a word.
Most platforms ask you to “describe your story” and start generating. FireQuill makes you do the structural work first, but makes it elegant. Eight layers. Booker’s seven plot archetypes, plus twelve additional narrative frameworks (Hero’s Journey, Heroine’s Journey, Time Loop, Save the Cat, Snowflake, more). Layered Jungian-inspired character archetypes. Genre-aware tone calibration.
Or use Holistic Setup: type a one-sentence pitch like “Mystery: a detective returns to her hometown to investigate a missing teen and uncovers her own buried family secret.” FireQuill drafts the full bible (frame, premise, themes, three to five characters with archetype plus want plus need plus flaw) in a single call. You review. You edit. Locked fields stay yours.
FireQuill is built on the bet that the work of writing is structural design, not typing. We make you do the design work, eight layers of it, before the typing starts. The agent isn’t replacing the craft. It’s amplifying it.
Auto-Flow runs the chapter agent in a loop, validates every chapter through the editor specialists, accumulates drift, pauses on compound failure or budget breach. You wake up to a draft.
Press the autogen button. The chapter agent reads your bible plus outline plus prior chapters plus recent character updates and writes Chapter One. Inline validators fire (line, developmental, continuity for fiction; argument, evidence for non-fiction). If verdicts are clean, the chapter ships and the loop advances. If a single specialist flags a warning, the chapter regenerates with corrective context. If multiple flags compound, the run pauses at a checkpoint and emails you to review.
Every chapter writes to your manuscript chunk. You can roll back any chapter, or all of them. A cost ceiling stops runaway spend. A drift budget catches voice, arc, and plot drift across the full book.
Auto-generate paragraph by paragraph and pause at every creative fork. The agent offers three to four directions; you pick the one that lands. Or write your own.
Auto-flow generation is unattended; Director’s Mode is the opposite: maximum agency. The chapter agent writes a paragraph, hits a decision point, and surfaces three to four trajectories: short labels, summaries, and 100-word in-voice previews. Sarah confronts him in the rain. Sarah lets him walk away, and follows. Sarah’s brother arrives, changing the calculus.
You pick. The book commits. Or type your own direction in plain English and the agent honors it. Repeat fork by fork.
Every word in your draft traces back to a decision you made. You stay the author. The agent stays the typist.
Fiction gets the character engine, plot validators, and scene-level extraction. Non-fiction gets the expert panel, argument validators, and section-level extraction. Same depth. Same care. Different bones.
Most platforms half-heartedly support non-fiction as an afterthought. FireQuill was built track-aware from layer zero. When you pick fiction at the wizard’s first step, the cast feature appears, the plot and continuity validators activate, and the wizard asks for genre, plot archetype, and comps. When you pick non-fiction, the panel feature appears, the argument, evidence, counterargument, and audience-fit validators activate, and the wizard asks for sub-format, reading level, scope facets, and citation style.
The engine, the agents, the snapshots, the auto-generation: all the same shape. The vocabulary changes; the depth doesn’t.
Genre, plot archetype, narrative framework, character archetypes, comps. Scene-level extraction.
Sub-format, reading level, scope facets, citation style. Section-level extraction.
Sign-up takes thirty seconds. The wizard takes fifteen minutes. The first chapter takes an hour, or a night, or however long the work of writing actually takes.
Start Writing for Free →