Release notes
0.2.0 — Easy Writer is now Ordfoss
August 2026
A new name
Easy Writer is now Ordfoss — Icelandic orð (word) and foss (waterfall). Same app, same team, same licence. Your existing projects, settings, window layout and preferences all carry over the first time you open it; nothing needs moving and nothing needs re-entering.
- Your existing files keep working. Projects saved as
.ewriterstill open exactly as before, and still open by double-clicking. New projects are saved as.ordfoss. Saving an older project leaves it where it is, under the name it already has. - One thing needs your attention if you use Claude. The built-in MCP server
is now called
ordfossinstead ofeasywriter. Open your Claude Desktop configuration and change that one word, or the Ordfoss tools will stop appearing. Help ▸ Connect Claude Desktop shows the exact snippet to paste.
The codex: properties are inherited now, not shared
A property used to belong to a whole category whether you meant it to or not: adding one while a single entity was selected gave it to every entity in that category, and to everything in the categories beneath. Properties are now defined where they belong and inherit downwards from there, so a codex holds what you actually put in it instead of accumulating empty slots on everything.
- A property belongs to what you put it on — adding a property while an entity is selected now adds it to that entity alone. Adding one to a category still gives it to every entity in that category and in the categories beneath. Each property says which of the two it is, so you can see whether an edit changes one thing or all of them.
- A category can sit under a single entity — a category can belong to one entity rather than to the project as a whole, and appears under it in the codex with its own properties. It inherits nothing from above by default; you tick which of the properties above it should take.
- A value that never varies is stored once — mark a property fixed and its value is held on the category. Every entity beneath shows that value read-only, and changing it on the category changes it for all of them at once.
- A sub-category can drop a property it does not need — untick any inherited property and it stops appearing on that sub-category and on everything beneath it. Nothing is deleted: values already entered are kept, and reappear if you tick it again.
- A property that links to something says what kind — a property that points at another entry can name the category it points into. Its list then offers only what is in that category and the categories beneath it, instead of everything in the project. Anything already linked stays linked and stays visible, whatever category it is in.
- A tidy-up for older projects — opening a project made before this offers, once, to move properties only one entity ever used down onto that entity. It backs up first, it never changes a value, and one Undo puts it all back.
- Decide the rest yourself — right-click the codex and choose Tidy Up Properties for the properties that automatic pass leaves alone: the ones no more than one entity uses. Each can be removed, kept on a category you pick from the chain above it, or handed to a single entity. Nothing is chosen for you, anything you skip is left as it is, and one tick applies your next choice to every remaining property.
- Categories split in two are put back together — the same window finds categories that are one category written two ways and offers to merge them, saying how many entities each holds. The one holding more is kept, gains every property the other declared, and takes the rest of the entities and sub-categories with it.
Write your language in its own letters
Constructed languages could be spoken but not written. A new Script window lets you draw the alphabet itself — by hand, by tracing something you already have, or by deriving letter shapes from how each sound is made — and then set your own text in it.
- Draw your own alphabet — freehand, lines, rectangles, ovals, polygons, stars, arcs and a pen tool, with snapping to guides and a grid, and a symmetry mode that draws the other half of a symbol as you go. Everything you draw stays editable point by point, whichever tool made it.
- It knows what you still have to draw — the sounds, syllables, roots or words your language already holds are listed for you, each marked drawn or not, and you pick the next one without leaving the canvas.
- Trace a drawing you already have — load a scan or a photograph of a page of letters and turn it into strokes you can then adjust.
- Derive letters from how a sound is made — draw the articulation of a sound as free lines, and where those lines cross is turned into a letter shape you can then edit like any other.
- Reshape letters you have already drawn — flip, rotate, scale, slant and shear, line points up, spread them evenly, or even out the spacing along a stroke. Apply it to the strokes you have selected, to the letter you are on, or to every letter at once.
- Rule your own guide lines — give a script a baseline, a headline, a centre line or anything else, say where each one sits, and have what you draw snap to it, to the grid, or to a fixed angle. The grid, the guides and the rulers can each be shown or hidden while you work.
- A writing system can be built from parts — a script whose symbols are assembled from a small set of shapes can describe those shapes and how they combine, instead of drawing every symbol by hand. A symbol can also record a flipped or turned version of itself rather than being drawn twice.
- See a passage in your own script — type any text and read it back in the letters you have drawn, then export the script, a specimen sheet, or a single letter.
Languages that carry more meaning
A word rarely means one thing in one way. This release lets a language record the readings, reversals and arrangements it actually uses, and the reference grammar reports all of it instead of a flattened summary.
- A word can mean different things in different places — a word whose meaning depends on where it sits, or on which sense it is carrying, can now record each of those readings on the one entry instead of being forced to pick a single definition and leave the rest in a note. Breaking a word down reports the reading that fits the position it landed in, and translating any of the readings finds the word.
- An accent that flips a word's meaning is understood — a language can say which mark it writes an opposite meaning with, and from then on a marked word is read as that word's opposite rather than as something unrecognised. Translating an opposite meaning writes the mark back on, so a word and its opposite no longer come out looking identical.
- A meaning can depend on an arrangement — a word that only means something when it brackets another, or when two of the same marker sit side by side, can now record that, and reading it back gives the combined meaning instead of the pieces.
- Grammar can describe the inside of a word — languages that build a whole sentence as one word can lay out the slots that word is made of, and that layout appears in the reference grammar.
- A much fuller reference grammar — the dictionary now shows what each word is built from, every meaning it takes, and where it came from, grouped by subject. It also no longer states settings you never chose.
Elsewhere
- Lock a document when it is done — right-click it and choose Lock. It stays readable and can still be reordered, but it cannot be edited or deleted, and the assistant cannot change it either. Only you can unlock it.
0.1.15 — languages that hold their own rules
July 2026
- Built-up words are understood — a language whose words are made by joining roots now has those words broken down and glossed root by root, instead of coming back whole and unexplained. A word can be marked as built from particular roots, and a root can carry a grammatical feature such as a tense, which now reads as that feature.
- One word, two opposite meanings — a word can record its opposite sense, so a language that flips a meaning rather than using a separate word can say so, and translating either sense finds it. Letters can do the same: a symbol can be recorded as another symbol turned around, rather than drawn twice.
- Edit or delete a single dictionary entry — correcting one definition no longer means resending the whole dictionary, which quietly dropped anything left out.
- Reading a language back no longer cuts off — grammar notes and dictionary entries can be read in full, and the summary now shows the syllable shapes the generator actually uses rather than a simpler stand-in.
0.1.14 — page view fixes, and an assistant that can find things
July 2026
- Pressing Enter no longer starts a new page — splitting a paragraph that began a page handed the second half its own page break, so a heading kept a page to itself and the text after your cursor was thrown onto the next one. A page break now stays on the paragraph you put it on.
- Changing alignment no longer breaks page view — centring or justifying a line dropped the page layout until you switched documents and back. The alignment buttons also behave as one control now, instead of leaving the previous one lit.
- The assistant can find what is in your world — searching through Claude now looks inside character and place fields, synopses, points of view, task and event descriptions, comments and plot threads, and tells you exactly where each match sits. It could previously only see names and chapter text, so a fact recorded on a character was invisible to it.
- Renaming reaches everywhere — the assistant can rename a character or place across your whole project in one step, and show you every place it would change first. It can also warn you when a new name is confusably close to one you already use.
0.1.13 — real pages, and exports you can aim
July 2026
- Page view now has real pages — text flows from one sheet to the next the way it will in print, breaking mid-paragraph where a book would, instead of a line being drawn across a paragraph that carried on past it. Each page is its own sheet with a numbered footer, so you can see how your book actually falls.
- Insert ▸ Page Break — start the next text on a fresh page. It shows up immediately in page view, survives saving, and is honoured when you export to PDF, Word, EPUB or standard manuscript format.
- Export just one part of the manuscript — the export window now has a “Part” picker, so you can export a single book, chapter or section instead of everything. Previously the only filter was by audience, which could not narrow anything down on its own.
- Set a default audience, and apply it to everything you have already written — anything you leave unrestricted is visible to every audience, so a “readers” export used to include your notes as well. Manage Audiences can now set what new content starts out visible to, and apply that across an existing project in one step, so you tag what readers should see rather than everything they should not. It leaves anything you have already restricted alone unless you ask it not to, and one undo reverses the whole sweep.
0.1.12 — important fix for cloud-synced projects
July 2026
- A project kept in a synced folder could lose entries — please update — if Ordfoss read your project file while OneDrive, Dropbox or iCloud was still downloading it, the parts that had not arrived yet looked as though they had been deleted, and Ordfoss removed them to match. It now recognises a file it cannot fully read and leaves it alone until the sync finishes. If you have lost anything, check File ▸ Back Up Project Now's folder — Ordfoss keeps automatic snapshots and your work is very likely recoverable from one.
0.1.11 — your work stays put
July 2026
- Changes made outside the app are no longer undone — when something else edited your project while it was open — the assistant working through Claude, a co-author, a sync client — Ordfoss picked the change up everywhere except the one document you had open, which quietly reverted to the text on your screen. Saving then wrote that older text back over the change. Both halves of that are fixed.
- Ordfoss will not save over a project that changed on disk — if the files have moved ahead of what you have open, saving now stops and offers to show you what changed instead of overwriting it. Nothing you have written is lost while you decide.
0.1.10 — fix it where you see it
July 2026
- Right-click a highlighted word to fix it — the coloured underlines told you something was wrong and then offered nothing. Now a right-click names the problem and fixes it: cut a filler word or a weak adverb, swap a wordy phrase for the short version, drop a repeated word. Cutting a word tidies up after itself, closing the gap and keeping the sentence's capital letter where it belongs. Where there is no single right edit — a passive sentence, one that runs too long — the assistant will rewrite it for you instead. You can also silence one phrase, or a whole kind of highlight, straight from the same menu.
- Contractions and possessives are no longer marked as misspellings — the dictionary held no words with apostrophes, so "couldn't" and "the woman's name" were underlined in red throughout ordinary writing, as were quoted words. Spelling suggestions can now correct a missing apostrophe too, turning "couldnt" into "couldn't".
0.1.9 — properties you can read
July 2026
- Every property is its own panel now — its name and buttons sit above the value instead of beside it, so a long summary uses the full width of the panel rather than a narrow column with empty space next to it. Fold any property down to a single line — or all of them at once — and each one still shows its value, so you can take in a whole character at a glance. The document panel works the same way.
- Drag a property to move it — pick one up by its name and drop it where you want it; a line shows where it will land. The up and down arrows are still there.
- Fields you inherited say where they came from — a field that belongs to a parent category is now labelled with it, and can be reordered or removed like any other one; removing it asks first, because it leaves that parent category too. Previously those fields quietly had no buttons at all.
0.1.8 — menus you can read
July 2026
- A new View menu, and panels grouped by what they do — the Windows menu listed all forty-two panels in one alphabetical run, with the layout commands buried among them. Those commands now have their own View menu — reset the layout, the writing modes, page and Markdown views, zoom, perspectives, theme and language — and Windows holds only the panels, sorted into Writing, World, Planning, Research, Progress, Data, Tabletop and Assistant.
- Update checks happen each time you start — previously at most once a day, so a fix could sit unnoticed for a day after you had already restarted. It is still a single request that sends nothing about you, and still switched off in Preferences ▸ General if you prefer.
0.1.7 — Preferences crash fixed
July 2026
- Edit ▸ Preferences no longer crashes — opening Preferences closed the program immediately, losing anything unsaved. It affected every user of 0.1.5 and 0.1.6, so if you are on either of those, please update. Nothing else in this release.
0.1.6 — the assistant can read your project
July 2026
- Ask about your actual writing — the built-in assistant can now read your manuscript, not just your codex. Ask it to summarise your chapters, check a scene against your timeline, or look something up in a spreadsheet or map, and it will open the real thing and answer from it. Previously it could only see codex entities, so a question about your documents got a guess.
- Answers without giving it permission to edit — reading now works with “Allow the assistant to make changes” switched off, so you can ask questions freely. Turn it on only when you actually want the assistant to write something.
- It remembers the conversation — you can correct it or ask a follow-up and it will take account of what you already said. Each request used to start from scratch.
- Clear errors from your AI provider — if a request fails, you now see what your provider actually said — a bad API key, an unknown model name — instead of “malformed response”. A spinner also shows while the assistant is working.
If you connect Google AI Studio, note that Gemini 1.5 model names no longer work with new API
keys — use gemini-2.5-flash — and leave the Speech URL empty, as Google's
OpenAI-compatible service does not offer voice generation.
0.1.5 — a clearer workspace
July 2026
- A new default layout — Ordfoss now opens as three clear columns: your manuscript and codex on the left, the editor in the middle, document details and the inspector on the right. The specialist tools stay closed until you want them, so every tab is readable. Nothing was removed — every panel is still in the Windows menu, one click away. If you have arranged your own workspace it is kept, and the program asks once whether you would like to try the new one.
- Custom themes — create your own colour theme, edit it, and share the file with someone else, alongside the built-in Light, Dark, Sepia and High Contrast themes.
- Much faster in the slow places — large spreadsheets, sorting by a calculated column, the relationship graph on a big codex, the screenplay preview, and Quick Open are all effectively instant now. Saving writes only what changed, so autosave no longer pauses on a large project.
- Word counts count prose — counts used to include formatting marks, so a chapter read longer than it was. Your word counts, readability scores and writing log will read slightly lower after updating, and they are now correct.
0.1.4 — working live alongside Claude
July 2026
This release smooths out using Ordfoss side by side with Claude and the built-in MCP server, makes the notifications that matter impossible to miss, and rounds off a few rough edges.
- Edits appear live, even while you work — when Claude edits your project through the local server, the change now shows up in the open app on its own even if you have unsaved edits of your own, as long as the two don’t touch the same thing. Previously, simply browsing the timeline (expanding or collapsing a lane) counted as an unsaved change and made every edit wait behind an “External changes” prompt. Now only a genuine conflict — you and Claude editing the same object — asks you to review.
- Notifications you can’t miss — an external change that conflicts with your unsaved edits, and an available update, now pop up a window instead of only a small indicator in the corner. A persistent indicator still remains if you’d rather deal with it later.
- File ▸ Exit — the File menu now has an Exit item at the bottom (it still prompts to save your work on the way out).
- For extension authors — the website now carries a complete, browsable C++ developer API reference: every public type and function you can build a native plugin on, grouped by area, with a getting-started guide for adding your own dockable windows.
0.1.3 — languages, relationships, and Markdown
July 2026
This release adds a full constructed-language toolkit, rebuilds the relationship graph so it stays readable at any scale, and moves your prose to a clean, version-control-friendly format — with no change to how the editor feels day to day.
- Build a real language — the Languages workspace is now a complete conlang designer: define your sounds, the rules for how they combine into words, a spelling system, and sound changes that let you derive daughter languages and dialects; keep a dictionary with parts of speech and etymology; set up grammar, paradigms, pronouns and numbers; gloss and translate passages; design a written script with a live preview; and export a formatted reference grammar.
- Relationship graph, rebuilt — a whole campaign’s worth of connections used to collapse into an unreadable wheel. The new graph is modelled on a reference viewer: focus one character and walk their neighbourhood a step at a time, filter by category or relationship type, colour nodes by category, search by name or alias, and highlight the shortest chain of relationships between any two entities. Your view settings are remembered per project.
- Cleaner, safer prose storage — documents are now stored as diff-friendly Markdown, so your work reads clearly in Perforce or Git and merges cleanly. An optional Markdown source view (Windows ▸ Markdown Source Mode) lets you edit the raw Markdown with syntax highlighting and a live preview; the normal formatted editor is unchanged and stays the default. Older projects are upgraded automatically and losslessly — nothing you have written is altered or dropped.
- Documentation & extensibility — a full user handbook and an extensibility reference (the plugin, scripting, and automation APIs) are now published on the site. Scripts can reach your whole project and add their own commands, and native plugins can contribute entire dockable windows.
0.1.2 — live sync with Claude, and total coverage
July 2026
This release deepens the bridge between Ordfoss and Claude: the assistant can now reach every part of your project, and the changes it makes show up in the open app the moment it makes them.
- Changes appear live — when Claude Desktop edits your project through the bundled local server, the open app now picks the changes up on its own, with no “External changes — Review…” step to click through. And the two never step on each other: the server re-reads the project before each change, so nothing you did in the app is overwritten.
- Every feature is reachable — the local server now covers the whole app: charts, typed and multiple-choice codex fields, continuity rules, footnotes, alternate “what-if” drafts, the submissions tracker, screenplay export to Fountain and Final Draft, project backups, and more — so you can ask Claude to build or edit anything, and each request is still a single undo step.
- Editor & inspector polish — inspector fields grow to fit their contents so a long value no longer truncates; references in the inspector are clickable links; the editor has a right-click spelling menu with a personal dictionary; and you can silence an individual style suggestion you don’t want.
0.1.1 — AI integration & single-instance
July 2026
This release makes the AI a first-class collaborator that can actually build your world with you, adds a way to drive Ordfoss from a Claude subscription, and smooths out how projects open.
- Convert a folder with AI — File ▸ Import ▸ Convert a Folder with AI… turns an entire folder of ordinary files (prose, images, maps, spreadsheets, PDFs, notes, bibliographies, stat blocks, and more) into a complete project: entities, timeline, sheets, maps, sources, and tasks, all cross-linked. An autonomous mode converts everything with no questions; a supervised mode reviews each stage.
- The assistant can make changes — turn on “Allow the assistant to make changes” in AI Chat and ask it to create characters, draft a chapter, or fill in the timeline; every request applies as a single undo step, so one Ctrl+Z takes it all back.
- Use your Claude subscription — a bundled local server (Help ▸ Connect Claude Desktop…) lets Claude Desktop, on the plan you already pay for, open and operate your project directly, with no API key. Every Ordfoss feature is reachable — including turning a PDF or Word manuscript into a full project with characters, places, and a timeline — and changes save immediately; if the project is open in the app too, it offers to merge.
- Single-instance file opening — opening a project when Ordfoss is already running now asks whether to use the current window or a new one, the way an IDE or word processor does, instead of quietly launching another copy.
0.1.0 — early-access preview
July 2026 · first public build
The first preview release, and it arrives with the whole workshop rather than a teaser:
- Single-file projects — a project is one
.ewriterfile you double-click, copy, and back up; or a plain folder of text files for version-control workflows. Atomic saves, automatic crash recovery, and clean conversion both ways. - Writing — full rich-text editor with styles, tables, footnotes, and embeds; focus, typewriter, and distraction-free modes; page layout view; project-wide find & replace with regex; autocorrect and smart punctuation.
- Worldbuilding — the codex with custom categories and fields, automatic name and alias linking in prose, entity galleries, merge/split, state-over-time, and user-defined continuity rules.
- The validating timeline — custom calendars, eras, lifespans, uncertainty, relative constraints, and multiple reckonings, with conflict detection for two-places-at-once, lifespan violations, and impossible travel.
- Maps — regions, routes, pins, tokens, battle grids, progressive reveal, and travel validated against the timeline.
- Screenplay mode — Fountain editing, industry pagination, coloured revision pages, Final Draft round-trip, and production breakdowns.
- Research — sources with APA/MLA/Chicago citations, BibTeX import, an in-app PDF reader with quote capture.
- Spreadsheets & charts — a real formula engine, cell formatting, CSV round-trip, embeddable charts.
- Planning & analytics — plot threads, beat sheets, corkboard, kanban, submissions tracker, goals, sprints, pacing and readability analytics, exportable health reports.
- Import/export — imports from Word, Scrivener, yWriter, Obsidian, Notion, Manuskript, and plain text; exports to DOCX, EPUB 3, PDF, standard manuscript RTF, Markdown, Fountain/.fdx, and a searchable static website.
- Bring-your-own AI — chat, inline assists, canon-consistency checking, semantic search, agents, and image/speech generation against any OpenAI-compatible endpoint you configure, with per-object opt-out and cloud redaction.
- Extensibility — JavaScript macro console, C++ plugin API, custom themes, rebindable shortcuts and toolbars.
- Languages — English and a complete Spanish translation.
Earlier internal milestones aren't listed here; public releases will be.