Write your novel and build its world in one place. Every scene, character and secret stays linked, your co-writers write alongside you, and Plotra keeps track of canon so you don't have to.
Try it free · Open the demo · Self-host it · Read the docs · Join the Discord
Think of Obsidian built for stories, with more than one person in it. The manuscript and the world live in one linked project: scenes, characters, places, a timeline, a plot grid. You write together live, and three things no other writing tool has together keep the story straight:
- The spoiler horizon. Tag a note, or a single paragraph, with the book or episode that reveals it. Set "World as of Book 2" and the whole project shows only what the audience knows by then: the binder dims the rest, the editor veils unrevealed passages, and the graph, timeline, codex and exports are filtered. Write a recap, a fan wiki or the next book without spoiling it. For a game master, a wiki exported at the horizon is the player-safe version of the campaign.
- Who knows what. Record which character learns which secret, and in which scene. It catches "how does she know that?" before a reader does.
- Continuity checks. Plain rules, no AI: someone appears after their death, an age doesn't match the timeline, a character is in two places at once, a closed plot thread comes back, a payoff lands before its setup.
Try it on a sample project without an account: the fantasy novel The Ninth Bell or Romeo and Juliet. Either can be copied into a workspace of your own.
| Graph | Timeline |
|---|---|
![]() |
![]() |
| What it is | Cost | |
|---|---|---|
| Plotra Cloud | The hosted version. Sign up and start writing; share projects and write with others. | Free plan, or Pro at $10 a month per workspace |
| Desktop app (coming soon) | macOS, Windows and Linux. Use it as a window on Plotra Cloud, or keep a library in a folder on your computer: no account, no internet, Markdown copies and snapshots beside it. | Free |
| Phone app (coming soon) | iPhone and Android. Read and write your projects on the go, catch ideas in a quick note, see mentions and reviews. Works offline and sends changes when you're back. | Free |
| Self-hosted | Your own server, with one docker compose command: the app, Postgres, the collab server and file storage. |
Free, no limits you didn't set |
Whichever you pick, you can export everything: a whole project as a .plotra file that opens in any of the three, or as Markdown, DOCX, EPUB and Fountain.
Writing and organising
- A workspace like Obsidian's: a binder of folders, chapters, scenes and notes, tabs and split panes, an inspector, a command palette (
⌘P), a quick switcher (⌘O) and hotkeys you can rebind. - A rich-text editor with headings, lists, tables, callouts, images, a slash menu and a floating toolbar. Focus mode and typewriter scrolling.
- Wiki-links (
[[Note]],[[Note#Heading]],[[Note|alias]]), embeds, hover previews, backlinks and unlinked mentions. - Note types with your own fields, templates, tags and saved layouts.
- Presets that make it feel like Scrivener, Plottr, Google Docs, Obsidian, World Anvil or a distraction-free writer, and settings for nearly everything else.
Seeing the story
- Graph, codex, outline, corkboard, plot grid, relationship map, timeline with custom calendars, storyboard canvas, manuscript, and a paginated read and print view.
Writing together
- Live co-editing with named cursors. Edits made offline are kept and merged when you reconnect.
- Comments and replies, suggestions (track changes), and version history with named snapshots, a diff and restore.
- Workspaces with members, and project sharing by email as viewer, commenter or editor, including people outside the workspace.
Story worlds
- A world bible with your own entity types and fields, shared between projects.
- Canon status on notes and single passages, with retcon history.
- Releases (books, episodes, sessions), the release roadmap and the spoiler horizon.
- Who-knows-what, setups and payoffs, and continuity checks.
Scripts and modules
- Screenplay, stage play, comic and audio formats, with Fountain import and export.
- Modules for research and citations, editorial review, production breakdowns, branching dialogue and quests, localization, and publishing (EPUB, a series bible, a spoiler-safe wiki).
AI, if you want it
- An agent that can read, outline and edit with you, on your own API key (Anthropic, OpenAI, OpenRouter, any OpenAI-compatible URL) or a local model through Ollama. It runs with your access and no more, and is off until a workspace owner turns it on. What you write is never used to train AI.
Self-hosting is free, and so will the desktop app be. The hosted version is priced per workspace, not per person.
| Free | Pro | |
|---|---|---|
| Price | $0 | $10 a month, or $96 a year |
| Projects per workspace | 10 | Unlimited |
| Storage per workspace | 50 MB | 50 GB |
| One file | 10 MB | 100 MB |
| Workspace members | 3, counting the owner | 25 |
| Guests per shared project | 5 | 200 |
| Features and modules | All of them | All of them |
| AI agent | Your own key or a local model | The same, plus hosted AI: 1,500 credits a month |
- No feature is locked behind Pro. Pro is higher limits, more collaborators and hosted AI.
- Going back to Free deletes nothing. The workspace keeps its data and only can't grow past the Free limits. Export always works.
- Payments go through Creem as merchant of record. Cancel any time from Workspace settings → Plan & billing; Pro runs to the end of the paid period.
- Hosted AI: one credit is 1,000 input tokens, output counts five times, and credits reset each billing period. Your own key still works on Pro and uses no credits.
- Smaller limits, the same on both plans: the latest 40 automatic versions are kept per note (named versions are always kept), and a person can send 40 invitations an hour. Self-hosters set or lift every limit.
If Plotra is useful to you, you can also sponsor it on GitHub. Sponsoring buys nothing extra; it pays for the Free plan and the work on the software.
Plotra runs on your own server with Docker, or on your own Postgres. The self-hosting guide in the docs covers setup, configuration and updates. To work on the code, start with CONTRIBUTING.md.
| Concern | Choice |
|---|---|
| Framework | Next.js 16 (App Router), React 19, TypeScript, Bun |
| Monorepo | Bun workspaces + Turborepo |
| Styling and UI | Tailwind CSS v4, shadcn/ui |
| Icons | Solar, through Iconify, loaded offline (@plotra/icons) |
| Editor | Plate.js |
| Real-time | Yjs with @platejs/yjs; y-indexeddb for the offline cache |
| Collab server | Cloudflare Workers + Durable Objects with y-partyserver, one Durable Object per document |
| Database | Postgres (Neon when hosted) with Drizzle ORM and row-level security |
| Auth | Better Auth: email and password, Google, magic link; its organization plugin for workspaces and its JWT plugin for collab tokens |
| File storage | Cloudflare R2, or any S3-compatible store |
| SMTP (any mail service) | |
| AI agent | Vercel AI SDK, on the user's own key (Anthropic, OpenAI, OpenRouter, any OpenAI-compatible URL) or a local model (Ollama) |
| Payments | Creem, as merchant of record |
| Panes and tabs | dockview |
| Command palette | cmdk |
| Graph | sigma.js + graphology |
| Canvas | React Flow (@xyflow/react) |
| Timeline | Custom, on dnd-kit |
| Docs site | Fumadocs |
| Hosting | Vercel (app, site, docs), Cloudflare (collab server) |
┌──────────────────────────┐ HTTPS ┌──────────────────────┐
│ Next.js app (Vercel) │ ───────────────────▶ │ Postgres (Neon) │
│ - UI, Plate editor │ Drizzle │ users, workspaces, │
│ - Better Auth │ │ projects, nodes, │
│ - Server Actions / API │ │ comments, versions │
│ - issues collab JWT │ └──────────────────────┘
└───────────┬──────────────┘
│ WebSocket (Yjs sync + awareness), JWT
▼
┌──────────────────────────┐ ┌──────────────────────┐
│ Cloudflare Worker │ │ Cloudflare R2 │
│ y-partyserver │ │ snapshots, uploads, │
│ 1 Durable Object / doc │ │ exports, backups │
│ (SQLite-persisted Y.Doc)│ └──────────────────────┘
└──────────────────────────┘
- A person signs in and opens a note.
- The app checks their access with one server-side
can(action, project)check and mints a short-lived JWT carrying their role. - The editor connects to the note's room on the collab server, which verifies the token. Viewer and commenter connections are read-only.
- The Durable Object stores the Yjs document, and after a pause posts its text, word count and links back to the app, which indexes them in Postgres.
What lives where. Yjs holds what several people type into or drag at once: document text, storyboard canvases, branching dialogue. Postgres holds what gets queried: the binder tree, properties, links, relationships, plot threads, timeline events, comments, versions. Every text note is its own Yjs document.
Access. Workspace members are owner, admin or member. A project can be shared by email with anyone as viewer, commenter or editor, and can have one read-only link that opens without an account. There are no per-note permissions.
The AI agent has no permissions of its own. Its tools wrap the same server actions people use and run as the person who started the run. It is off until a workspace owner turns it on.
apps/
web/ the marketing site (Next.js)
app/ the product (Next.js)
collab/ the collab server (Cloudflare Worker)
docs/ the docs site (Fumadocs)
desktop/ the desktop app (Electron), cloud window or local library
packages/
ui/ @plotra/ui: shadcn components and the Tailwind theme
icons/ @plotra/icons: the Solar icon wrapper
db/ @plotra/db: Drizzle schema, migrations, client
auth/ @plotra/auth: Better Auth config and JWT helpers
config/ shared tsconfig and ESLint config
docker/ compose file and Dockerfiles for self-hosting
scripts/ backup.ts, check-licenses.ts
- Internal packages are consumed as TypeScript source, so there is no separate build step.
- Apps import only from
packages/*, never from each other. turbo run dev|build|lint|typecheckfrom the root.- Auth lives only in
apps/app; the site just links to sign-in and sign-up.
Issues, ideas and pull requests are welcome. Read CONTRIBUTING.md first: it covers setup, how the code is organised, and the few rules that matter (the main one: talk before you add a database migration). For anything larger than a small fix, open an issue before you start.
- Discord: questions, ideas, and talking a change through before you build it
- Code of conduct
- Reporting a security problem
Plotra is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0-or-later). If you run a modified version as a network service, you must make your source code available to its users.
Icons are Solar by 480 Design, used under CC BY 4.0. The licences of the other dependencies are checked by bun scripts/check-licenses.ts.



