FAQ
The app UI supports Chinese and English (switch in Settings). This site is available in both languages.
Install and setup
How do I install All Writer?
See Installation. Short version:
- Download from Download
- Run the installer (handle SmartScreen if prompted)
- Finish the first-run wizard
Or use the web app with no install.
SmartScreen blocks the installer
Click More info → Run anyway. A code-signing certificate is planned.
Which platforms?
- Desktop: Windows 10+ (x64)
- Web: modern browsers; mobile browsers work as a responsive web app (no separate native app)
macOS / Linux desktop builds are on the roadmap.
Uninstall (Windows)
Settings → Apps → All Writer → Uninstall. Your document data stays under your user folder unless you delete it manually.
Data and sync
Where is data stored?
- Desktop: local database by default.
- Web: cloud storage with your account.
Cloud sync on desktop?
The desktop app is local-first; full two-way cloud sync is not the default—check product updates.
Backup
Export a Markdown ZIP, AI share cards, or a Web package (Export). You can also rely on web cloud backup if you use the web app.
Real-time collaboration?
Not in the current release—planned for the future.
AI Agent
What can the Agent do?
One AI Agent for the whole project: outline, draft, edit with tools, @file / @web / @base, URL-to-Markdown, read-only subagents, custom Chat models, diff review, Mermaid diagrams, and more. Details: AI features.
Planning, writing, review?
planning— outline, structure, task breakdownwriting— draft, expand, rewritereview— fact/logic/style pass
Use plan → write → review in one thread when possible: Writing, Review. System extraction and AI document import are launched from dedicated product entry points, not from the everyday stage selector.
What are subagents?
Subagents are read-only helper runs in separate sub-conversations. explore_project maps structure and context, review_project reviews with evidence paths, and research_context gathers source-backed context. They do not directly edit your manuscript. If a reference report is saved, it appears under @system/@subagent/; actual manuscript edits still go through Diff review.
Can I use my own model API key?
Yes. Add an OpenAI-compatible Base URL, API key, and model ID in Settings → AI Features → Custom Chat models. Run the provider/model test before use; failed tests show redacted request/response details to help diagnose Base URL, key, or streaming Chat Completions issues.
Custom model bills are handled by your upstream provider. All Writer platform abilities such as web search, knowledge-base search, URL-to-Markdown, and AI Word export still consume All Writer balance when used.
How do @ commands work?
@filename— attach file context (picker after@).@web— web search, ¥0.01 per call (cached repeats usually avoid repeated charge).@base— semantic search over your project content.- URL-to-Markdown — paste a URL and ask the Agent to summarize, rewrite, or save useful points as Markdown context; ¥0.01 per call.
Does AI cost money?
Yes—usage draws from your CNY balance. See Pricing.
What is @system?
A special folder for project rules (style, glossary, characters, etc.) the Agent reads automatically. See @system.
Templates?
Everything in one place: Template creation.
Does AI need the network?
Yes for AI features. Editing, saving, and export on desktop can work offline without AI.
Do you store my manuscript?
AI features process relevant text only when you start an AI request. All Writer acts as an AI forwarding and orchestration service, so the context needed for the request is sent to the configured model gateway or upstream model provider. AI edits still go through diff review before they are applied; upstream handling follows the relevant provider or gateway policy. Conversations, task metadata, and service logs may retain the operational data needed to run the product.
AI not working?
Check network, account balance, login status, and any maintenance notices.
Edit a past chat message?
Click a message to edit; the app may ask to remove later messages before regenerating—useful to fix prompts.
@web vs @base
| @web | @base | URL-to-Markdown | |
|---|---|---|---|
| Source | Internet | Your project | A specific web page |
| Good for | Fresh facts | Lore, specs, internal text | Summarizing or reusing a page |
| Billing | Per policy + cache | Usually lower than long generations | Per actual call |
Editor
How do I know it saved?
Toolbar status + dot on unsaved files + auto-save shortly after edits + prompt on exit if dirty.
Markdown support?
Full common Markdown + tables, tasks, math (KaTeX), Mermaid, code fences—see Markdown and feature pages.
Recent builds also improve long-form rendering with Markdown block caching, streaming throttling, lazy Mermaid rendering, local tool-call error boundaries, and stable timeline ordering for streamed conversations.
Shortcuts?
Still stuck?
Email ypl123698745@qq.com or open a ticket on Community.