Where BookForge
is heading next.
Curated public view of what's next. Themes — not slot numbers, not commit hashes. Order in Later shifts as we learn what authors actually need most. If a theme matters to you, tell us and we'll re-weight.
The last three updates.
Every theme below moves toward this list over time. See the full changelog →
-
Hand off your manuscript to a human translator
**You can now export a copyedited volume as a portable .zip** — chapters with frozen markers preserved verbatim, your style sheet, a glossary built from your Reference Library and…
Read entry → -
Declarative pipeline shape — internal refactor (no user-visible changes)
**Internal refactor — no user-visible changes.** The pipeline definitions migrated from absolute `sequenceNumber: N` integers to a declarative `position: { after, before }`…
Read entry → -
Hotfix — accessibility apply steps (C10.7.A / C11.5.A) sidecar parse
Fixes a ship-blocker discovered in 1.26.5: the new accessibility apply steps C10.7.A (PDF/UA) and C11.5.A (EPUB) threw on the first sidecar parse. Cause: `'accessibility'` was…
Read entry →
Next up — actively being built
These ship in the next 1-2 weeks. Already in scope, already scheduled, no surprises.
-
Local-model support (Ollama)
Route prose-generation steps through Ollama on your machine instead of Claude. Editorial steps that need tool access still go to Claude. Free local inference, $0 per generated chapter.
Soon — scoped, in scheduled queue
These are scoped and queued. Each is ~1-3 weeks of work, shipping after the items above.
-
DOCX round-trip + editorial scorecard
Receive a manuscript as DOCX, run the editorial pipeline against it, return DOCX with reviewer comments tracked + a quantitative scorecard. The trade-publisher workflow.
-
Bibliography + style sheet auto-generation
CSL-formatted bibliography pulled from your References library + a descriptive style sheet (your own house rules, captured from the manuscript) on every Final Edition pass.
-
Print-spec compliance lint
Validate your PDF against KDP, IngramSpark, and Barnes & Noble print specs before you upload. Catch trim, bleed, and margin issues at compile, not at rejection.
-
Index, glossary, dramatis personae
Deterministic back-of-book generators that pull from your References library and manuscript text. No more hand-typing the index — page references stay accurate as the book evolves.
-
Front + back matter compose
Half-title, copyright page, dedication, epigraph, TOC, acknowledgments, about-the-author, also-by, colophon — generated from project metadata, ready to ship.
-
PDF/UA + EPUB Accessibility audits
Validate every export against WCAG and the accessibility-metadata fields readers and assistive tech need. Stay ahead of the EU Accessibility Act deadlines without lawyers.
-
Translator handoff package
Decouple the translation track from "AI-only translation". Bundle source text, reference glossary, voice profile, and style notes into a package a human translator can work from.
-
Authorship ledger
Per-paragraph provenance chain — who wrote it, whether AI assisted, what edits applied. The transparency story for publishers, agents, and readers who ask the question.
-
Eliminate the terminal sign-in step
Sign in to Claude entirely from inside BookForge. No more "open a terminal and type claude" instruction. One-click OAuth from the welcome screen.
-
Author ergonomics bundle
Editor focus mode (hide chrome, center serif typography, F11 toggle), one-click EPUB-of-one to send to a beta reader, project archive export, backup health indicator in the project header.
-
Auto-approve gates
Per-project policy: auto-approve the low-risk audit steps (Conservative / Balanced / Aggressive presets) and keep manual gates on the high-risk ones — legal, sensitivity, final assembly. Run-while-you-sleep.
-
Run-to-here button
Click any step in the matrix and have BookForge run every chapter from its current state to that target step. Hand-edit safeguards prevent clobbering manual changes.
-
Pipeline presets
Named profiles bundling model picks, gate decisions, and concurrency. Four built-in (Quick draft / Full production / Translation pass / Galley-only) plus your own saved profiles.
-
User Guide refresh + AI assistant
Comprehensive in-app User Guide rewrite, accessibility-correct from day one, with deep-linking and per-step sections. Plus a chat assistant inside the guide that answers questions about any pipeline step.
-
Planning workspace
New "Plan" tab between Discovery and Pipeline. Absorbs Outline and References, ships a persistent planning chat assistant, surfaces Plan Health — what's missing before you can start drafting.
-
Editor reframe
Markdown serif typography in the prose pane (readable measure, generous leading). Footer status bar with live word count, session goal, "saved Xs ago", reading time. Find-and-replace. Word-level prose diff that reads like Google Docs.
Later — committed direction
Committed direction without locked timing. Order will shift as we learn what authors actually need most.
-
Outline 2.0
Conversational outliner with per-beat anchors that live in the editor. Drag-and-drop chapter ordering. "Draft this beat" generation that respects the surrounding context.
-
Voice Monitor 2.0 + Pacing dashboard
Shared stylometrics, percentile-against-anchor scoring (vs your own best chapters), multilingual emotion lexicons, drift timeline per metric, reader-energy alternation warnings.
-
Visual fetch pipeline + Sourcing Review
Search-and-fetch interior illustrations, maps, and data visualizations with proper provenance and image-rights tracking. Workshop tool for per-opportunity preview / accept / reject.
-
Compare 2.0
Three preset comparison modes (current vs prior, draft vs finished, two variants) reachable from every relevant surface. Word-level prose diff that reads like Google Docs Suggesting.
-
Timeline 2.0
Status-coded bars, time axis, filter strip, rich tooltips with verdict + model + cost + triggered-by, apply-step pairing brackets, manual-override badges, CSV export.
-
Discovery 2.0
Per-phase model selection. Rewind and fork conversations. Multi-author Discovery with per-turn attribution. Live Brief gap analysis. Resume after disconnect.
-
Spoiler-aware references + Chat with character
Per-attribute "reveals at chapter X" filtering. Chat with one of your own characters — multi-mode (Haiku quick / Sonnet considered / Opus deep) — without leaking late-book information.
-
Beat-kind enrichment + reader energy
Expanded beat vocabulary (action / emotional / dialogue / revelation / decision for fiction; narrative / insight / transition for non-fiction). Chapter-rhythm scaffolds + alternation warnings.
-
Author safety + AI-detectability self-check
Per-project archive bundles, automatic backups, and a "how AI-detectable does my own writing look" self-mirror — never an accusation tool, always your own quality bar.
-
Citation + bibliography exports
CSL-JSON and BibTeX exports from the canonical-references graph. AI-suggested citations grounded in your source library — never invented, always traceable.
-
Mobile capture inbox
Capture-only PWA for jotting ideas and voice notes from your phone. Syncs to your active project's research drawer. No editing on mobile — capture only, by design.
-
Multi-window mode
Pop the Live Agent Scribe out to a second monitor while you write on the primary. Tear off the Editor for chapter-side-by-side work.
-
Article + periodical formats
Workshop tool for shorter pieces (Launch Companion). Periodical / serial / newsletter formats land later, gated on enough authors asking for them.
Already shipped
Already in your BookForge. Pull the latest build to use them.
-
Editor-in-control review gates
Every audit step lands in a Review Center where you approve, edit, or send back — per finding, with an audit trail. The pipeline waits on you, not the other way around.
-
Live agent scribe
Watch every AI step as it works — the prose it's writing, the files it's reading, the tools it's using, the cost as it ticks. Pause it. Copy the transcript. Open the raw log.
-
Pipeline Matrix
Chapters as rows, production steps as columns, status dots per cell. See the whole production state of your book at a glance; click any cell for step detail.
-
Discovery conversation
A guided nine-phase conversation walks you from a one-line idea to a complete book blueprint. Auto-progresses; you can rewind, fork, or take over the wheel at any point.
-
Reference Library
Your book's people, places, sources, claims, and concepts in one canonical store. Characters stay consistent, facts stay sourced, citations stay aligned across every chapter.
-
Per-step model picker + cost insights
Pick Haiku for cheap audits, Sonnet for the rest, Opus for hard drafts. Measured cost per step, cache hit rate, and monthly spend — surfaced where you make the call.
-
Multi-reviewer panel
Seven simulated reader perspectives — scholar, narrative, general, adversary, cultural, commercial, custom — read your manuscript in parallel and synthesize one apply-step.
-
Translation pipeline
Run the full editorial pipeline against a target-language draft. English, French, Spanish, German, Italian, Portuguese; voice supervised in the new language, not just translated.
-
Uncertainty-tag review
When the AI isn't sure about a fact it marks it with a tag. Resolve every open uncertainty from one drawer before publishing — no surprise [?] markers in the printed book.
-
Atomic save + crash recovery
The editor uses atomic writes plus a 2.5-second background autosave. Power outage mid-chapter? You pick up where you left off. Frozen prose stays frozen across edits.
-
Auto-install Claude
First-launch install button runs Anthropic's official installer for you and walks you through connecting your account. No terminal commands to copy-paste.
-
SHA-256 artifact provenance
Every chapter draft, every reviewer decision, every export gets hashed and registered. Trace any line in the final book back to the prompt, model, and timestamp it came from.
-
Final Edition print-PDF + EPUB + ONIX
Distribution-ready files: typeset print PDF with page labels and table of contents, validated EPUB, and ONIX 3.0 metadata package for retailers.
-
Your work stays on your machine
BookForge runs entirely on your computer. Manuscripts, drafts, and review reports never leave your disk. Claude is spawned locally against your own account.
-
Non-technical install polish
Auto-fix terminal PATH after install. Plain-English copy instead of "Claude Code CLI" jargon. Progress checklist with named milestones (Downloading / Verifying / Installing / Updating PATH) instead of streaming shell output. One-click "Test Claude connection" button in Settings + onboarding.
-
Plain-English step names
Pipeline steps now read "1 · Write", "2 · Developmental Edit", "3 · Fact-Check" — sequence numbers + plain names across the matrix, Step Detail, Review Center, Timeline, dialogs. Renamed Apply * Remarks → Edits, Adapt → Translate, Libel → Legal, Claim Resolution → Resolve Citations.
We don't ship features just because they're on a roadmap.
Every theme above earns its slot by being the next-best thing for authors actually using BookForge today. If a theme stalls in Later for months, it's because something else surfaced as more load-bearing — not because we forgot.
The changelog is the source of truth for what's actually shipped. The roadmap is the source of truth for where attention is going. A theme on this page matters to us? Tell us and we'll re-weight.