From discovery to distribution-ready files — orchestrated through one dashboard, powered by your own Claude account, with editor-in-control gates at every step.
BookForge dashboard · actual desktop app screenshot · not a render
BookForge runs every pipeline step through Claude. You bring your own Claude account — subscription or API key. We never touch your spend or stand between you and Anthropic.
Sign in once with your Pro or Max account. Pipeline steps run against your subscription's usage limits — predictable monthly cost, no per-token billing, no surprise invoices.
Drop in your own API key from console.anthropic.com. Pay per-token at Anthropic's published rates — better if you want fine-grained spend control or you're running short experiments.
Same product either way — every pipeline step, every gate, every export format works identically. Switch between billing modes from Settings → AI at any time.
Every audit step — fact-check, continuity, supervision, copyedit, galley proof, sensitivity, legal — lands in a Review Center. You approve, edit, or send back. The pipeline waits.
The AI proposes; you dispose. Every chapter draft is hashed and versioned. Every reviewer decision is logged. Every artifact is traceable back to the prompt, the model version, and the timestamp.
No black boxes. No surprises in the final manuscript. Just a carefully sequenced graph of independent steps, each one inspectable, each one revertable, each one yours to gate.
Run discovery, generate the blueprint, outline every chapter, draft, edit, copyedit, galley-proof, typeset, export EPUB + ONIX. Approve each gate as you go. Manuscript never leaves your machine.
Bring your English manuscript; pick a target language (French, Spanish, German, Italian, Portuguese). BookForge runs the translation track, humanizes the prose, supervises voice, and re-typesets for the new edition.
Skip the draft; bring a finished book and run the galley-proof gate. Receive structured remarks per chapter — continuity, fact-check, copyedit, sensitivity — and apply them with per-remark accept / reject control.
Seven reader perspectives — scholar, narrative, general, adversary, cultural, commercial, custom — read the whole book in parallel. Synthesize their notes into a single apply-step. Used before pitching to agents or hitting publish.
One author's manuscript — Artificial Wealth: The Intelligence Bubble — partway through the pipeline. Real prose, real gates, real metrics. Every screen below is an actual BookForge screenshot, not a render.
Your manuscripts never leave your machine.
BookForge runs entirely on your computer. The desktop app spawns Claude as a subprocess against your account; manuscripts, drafts, and review reports stay on disk. Every artifact is SHA-256 hashed and registered in a local SQLite database — so every line of the final book traces back to the prompt, the model version, the reviewer's decision, and the exact timestamp.
No telemetry, no cloud uploads of your work, no third-party intermediary. Read the privacy notice.