What a Document Collaboration Platform with AI actually does
Provides block-based document editing with AI-powered section drafting, real-time translation, and workspace Q&A over stored documents.
A document-collaboration platform is a rich-text editor (like Notion, Google Docs, Coda) that lets teams co-author wikis, playbooks, product specs, and knowledge bases. The core architecture: frontend (React/Vue block editor) + backend (CRDT for conflict-free edits + WebSocket sync) + storage (PostgreSQL for document structure + full-text search). Paragraph 2: Notion ($10B valuation), Google Docs (free, part of Workspace), Coda ($1B+ valuation), Confluence Cloud ($5B Atlassian), and Microsoft Loop all own this space with years of engineering polish. Greenfield is a 12–24 month, $300K–$1M+ commitment with a <5% win rate. The open-source alternative—AppFlowy (50K+ GitHub stars, Notion-clone), Outline (24K stars, wiki-first), HedgeDoc (Markdown-based)—gives you 80% of the Notion UX for free. The AI angle is the thin layer on top: section drafting from outline (Sonnet 4.6 + caching), translate-as-you-type (Gemini 3 Flash), document Q&A over workspace via RAG, structured-data extraction (turn notes into tables).
AI capabilities involved
Section drafting from outline
Real-time translation (100+ languages)
Document Q&A over workspace (RAG)
Embeddings for semantic search & clustering
Who uses this
- Knowledge-management consultancies selling branded Notion replacements to 5–20 SMB clients
- Internal-wiki specialists (legal, healthcare, SaaS) needing a GDPR-compliant self-hosted alternative to Confluence
- Industry-specific doc-tool consultancies (e.g., legal case-management, clinical-workflow) needing custom branching and automation
- Engineering teams wanting a private wiki (no vendor lock-in) with AI-assisted documentation
SaaS alternatives on the market
Real products you can sign up for today — with current 2026 pricing, honest pros and cons.
Notion
99% of teams. If you want a doc tool and don't care about white-label, buy Notion and ship in 1 day.
Free (unlimited blocks, 1 guest, 1 sync block import)
$10/user/mo (Plus)
$25/user/mo (Team), enterprise quote (Enterprise)
Pros
- +All-in-one: docs, databases, CRM, forms, AI copilot, templates in one ecosystem
- +Network effects: 40M+ users, thousands of templates, integrations, and community plugins
- +AI copilot: auto-complete, content generation, summarisation (Q1 2026 feature)
- +Mobile app and offline sync
Cons
- −No white-label or reseller tier—your customers always see 'Notion' branding
- −Pricing inflation: raised from $8→$10/user in 2024; expected to hit $12–$15/user by end of 2026
- −Collaboration can be slow on large documents (>10K blocks); real-time sync latency 2–5 sec
- −AI copilot is separate from the editor—not integrated into the writing UX
Coda
Data-heavy teams (product specs, financial models, roadmaps) who want formulas + real-time collaboration without switching to Excel
Free (unlimited pages, 1 doc, community support)
$10/user/mo (Pro)
$30/user/mo (Team), enterprise quote
Pros
- +Tight formula/database integration—better than Notion for data-driven docs
- +Pack API for advanced automation—more powerful than Notion's automations
- +Native multi-player real-time editing with low-latency CRDT
- +Flexible pricing: team/workspace tiers (not per-document like Notion)
Cons
- −No white-label—Coda branding is locked in
- −Smaller user base (5M vs. Notion's 40M) = fewer integrations and templates
- −AI features lag Notion's copilot roadmap
- −Learning curve: formulas and Pack API are powerful but have steep UX
AppFlowy (OSS, self-hosted)
Consultancies and agencies building white-label doc tools for customers; engineering teams wanting a self-hosted private wiki
Self-host (free OSS; 50K GitHub stars) or Cloud free tier (limited)
$0 (self-host) or $99/mo (Cloud Business)
Self-host, no vendor fees
Pros
- +True white-label: 100% rebrandable when self-hosted on your domain
- +Open source: modify the editor, add custom blocks, integrate with your own backend
- +No vendor lock-in: all data stays in your PostgreSQL database
- +Low self-host cost: $5/mo Hetzner VM handles 50–100 users
Cons
- −80% feature parity with Notion: missing AI, some template features, integrations require custom work
- −Maintenance burden: you manage security patches, upgrades, backups, GDPR compliance
- −Smaller community: fewer third-party integrations and templates vs. Notion ecosystem
- −UX polish: UI is clean but doesn't match Notion's polish (animations, edge-case handling)
Confluence Cloud
Atlassian shops (Jira + Confluence) for technical documentation and project wikis
Free (up to 10 users, limited features)
$5.16/user/mo (Standard, min 3 users)
$224/mo (Premium), enterprise quote
Pros
- +Integration with Jira—seamless for engineering teams already using Jira
- +Advanced permissions and audit logs (Standard tier+)
- +Mobile app with offline editing
- +Atlassian Marketplace has 2000+ plugins
Cons
- −No white-label; Atlassian branding is mandatory
- −Pricing is per-user: $5.16/user/mo adds up fast for large teams (100 users = $516/mo)
- −UI feels dated vs. Notion/Coda—less focus on UX elegance
- −AI features are minimal compared to Notion's roadmap
The AI stack
A document-collaboration platform with AI requires two layers: the editor backend (document storage, CRDT sync, version history) and the AI layer (section drafting, RAG, translation). The editor is best sourced from AppFlowy (OSS, 50K stars) rather than built from scratch. The AI layer is new: Sonnet 4.6 with caching for section drafting, Voyage embeddings for document Q&A, and Gemini 3 Flash for real-time translation.
Document Editor & Storage
Block-based rich-text editor (like Notion) with real-time co-editing, version history, and full-text search
AppFlowy (self-hosted OSS fork)
$5/mo Hetzner VM for 50–100 users; $0 software licenseConsultancies and agencies building white-label tools; engineering teams with <500 users
Outline (OSS wiki-first)
$5/mo VM; $0 softwareInternal wikis, playbooks, knowledge bases (not data-centric use cases)
HedgeDoc (OSS Markdown-based)
$5/mo VM; $0 softwareEngineering teams wanting a lightweight wiki; not suitable for non-technical users
Our pick: AppFlowy for white-label builds targeting Notion-like UX; Outline for wiki-first use cases (internal playbooks, knowledge bases). Self-host both on Hetzner for $5–$10/mo.
Foundation Model: Section Drafting
Auto-generate sections of a document from an outline, title, or context
Claude Sonnet 4.6 with prompt caching
$3/$15 per M tokens; caching reduces cost 90% on repeated callsBatch section drafting (user clicks 'Draft next section' and waits 10 sec)
Claude Opus 4.8
$5/$25 per M tokensHigh-stakes docs (legal contracts, technical specs, clinical workflows)
Gemini 3.1 Pro (2M context)
$2/$12 per M tokensSection drafting where you need workspace-wide context (reference 100K tokens of docs)
Our pick: Sonnet 4.6 with caching as default (90% quality, cost-effective at scale). Upgrade to Opus 4.8 for legal/healthcare/technical verticals where quality is non-negotiable.
Embeddings: Document Q&A & Search
Index documents as vectors; allow users to ask questions over the entire workspace ('What was decided in the Q3 board meeting?')
Voyage voyage-3.5
$0.06/M tokensWorkspace Q&A with RAG (retrieve relevant docs, then ask LLM to answer)
text-embedding-3-large
$0.13/M tokensHigh-precision retrieval (legal discovery, healthcare records, sensitive data)
text-embedding-3-small
$0.02/M tokensCost-sensitive use cases where <90% retrieval quality is acceptable
Our pick: Voyage voyage-3.5 ($0.06/M) for workspace Q&A with RAG. Store embeddings in pgvector (Supabase) for fast cosine similarity search.
Real-Time Translation
Translate documents and blocks on-the-fly into 100+ languages
Gemini 3 Flash
$0.075/$0.3 per M tokensHigh-volume translation of casual docs
Claude Sonnet 4.6
$3/$15 per M tokensNuanced translation (legal docs, marketing copy, brand voice consistency)
Our pick: Gemini 3 Flash for translate-as-you-type on high-volume blocks. Route legal/brand-critical docs through Sonnet 4.6 via user choice ('Premium translation').
Reference architecture
The pipeline: user creates a new document in AppFlowy → selects 'Draft section' → AI layer receives document context (title, outline, related docs from workspace) → Sonnet 4.6 with caching generates section text → appended to document. For Q&A: user opens 'Workspace Q&A' panel → Voyage embeddings retrieve top-5 relevant documents → Sonnet 4.6 reads retrieved docs + user question → generates answer with citations. For translation: user selects 'Translate to Spanish' → Gemini 3 Flash translates block-by-block → creates shadow document (original + translated side-by-side).
User creates document in AppFlowy editor with title and outline
AppFlowy frontend (React)Document stored in AppFlowy's PostgreSQL database with block structure. Outline stored as heading blocks (H2, H3) with content placeholders.
User clicks 'Draft section' button → AI layer fetches document context
Supabase Edge Function (Node.js)Query AppFlowy database for: current document title, outline (H2/H3 blocks), 5 most-similar documents from workspace (via Voyage embeddings). Build prompt: 'You are drafting a section of a document about {title}. Here is the outline: {outline}. Related docs for context: {related_docs}. Draft the next section after the last heading.'
Call Sonnet 4.6 with prompt caching (cache the outline + related docs)
Anthropic Claude API with prompt cachingInput: 40K tokens (context docs, outline, related pages). Cache this for 5-minute TTL (user drafts multiple sections). Output: 500–1000 token section text. Cost: $0.12 non-cached; $0.012 cached (90% savings).
Insert drafted section into document, show to user for review/edit
AppFlowy frontendNew block appended to document with 'AI-generated' badge. User can accept, edit, or delete. Accepted section is final.
For workspace Q&A: user asks a question in AI panel
Supabase Edge Function + Voyage embeddingsUser types 'What are our company values?' → convert question to embedding via Voyage → cosine similarity search in pgvector for top-5 documents → retrieve full text of those docs.
Call Sonnet 4.6 with retrieved docs as context
Anthropic Claude APIPrompt: 'Using these documents: {retrieved_docs}, answer the user question: {question}. Cite which document each answer comes from.' Output: 200–500 token answer with citations.
Estimated cost per request
~$0.012 per section draft (Sonnet 4.6 cached, ~40K tokens input, 1K output); ~$0.005 per workspace Q&A query (Voyage embedding + Sonnet 4.6 retrieval-augmented, ~10K tokens input, 500 output); ~$0.0003 per translated block (Gemini 3 Flash, ~1K tokens).
Cost calculator
Drag the sliders to model your actual usage. The numbers update in real time so you can stress-test economics before writing a single line of code.
This calculator models monthly cost to run a white-label document-collaboration platform (AppFlowy fork + AI layer) for mid-market customers. Assumptions: each customer has 5–30 users, 100–500 documents, and 50–200 AI requests/mo (section drafts + Q&A + translations). Deployment: self-hosted AppFlowy on Hetzner + Supabase for AI orchestration + Voyage embeddings.
Estimated monthly cost
$70.10
≈ $841 per year
Calculator notes
- AI request volume varies: SMB wiki users generate 2–5 requests/user/mo; product/engineering teams generate 10–20 requests/user/mo.
- Caching assumption: 80% of section-drafting requests re-use the same outline template (90% cache hit = 10x cost reduction on input tokens).
- AppFlowy host is shared: one Hetzner VM ($20/mo) handles all customers' document storage and real-time sync; scales to 500+ concurrent users.
- Not included: your labour (customer onboarding, support, feature development), domain registration, or SSL certificate (included in Hetzner).
Build it yourself with vibe-coding tools
A white-label doc-collaboration MVP can run in a weekend: AppFlowy in Docker on Hetzner ($5/mo) + Lovable AI sidecar panel that drafts sections and Q&A. You'll have a working multi-user wiki with AI-assisted writing. Not production-grade (no per-customer auth, limited compliance), but enough to test demand with 2–3 early customers.
Time to MVP
16–20 hours (one weekend + 4 hours Monday debugging)
Total cost to MVP
$25 Lovable Pro + ~$50 Sonnet API credits + $5 Hetzner VM (first month)
You'll need
Starter prompt
Build me a document-collaboration app with AppFlowy backend + AI sidecar. Here's what I need: 1. AppFlowy integration: I'll host AppFlowy in Docker on Hetzner ($5/mo). My Lovable app should embed AppFlowy via iframe (display the editor) and add an AI panel to the right side. 2. AI Panel with 3 tabs: a) 'Draft Section': user gives a title + outline (e.g., 'Introduction, Main Points, Conclusion'), clicks 'Draft', and Claude Sonnet 4.6 generates a 500-word section. Insert into the AppFlowy document. b) 'Ask Workspace': user types a question (e.g., 'What are our company values?'), and the AI retrieves related documents from AppFlowy + answers with citations. c) 'Translate': user selects a language and the app translates the current document block into that language using Gemini 3 Flash. 3. Authentication: Email + password, stored in Supabase. One user per workspace (MVP: no multi-user auth per-document). 4. Data flow: - Sonnet calls via Supabase Edge Function (server-side API key, not exposed to client). - For Q&A: fetch related documents from AppFlowy API, embed using Voyage, store in pgvector, retrieve on query. - All drafted sections stored in a separate 'drafts' table for version history. 5. UI: Split-pane layout — AppFlowy iframe on left (70%), AI panel on right (30%). Dark theme, Tailwind. Show token usage (cost) at the bottom of the AI panel ('~$0.01 per draft'). Used by: knowledge-management consultancies selling to SMB clients wanting a branded doc tool with AI.
Paste this into Lovable
Follow-up prompts (run in order)
- 1
Add a 'Workspace Settings' page where the user can configure AppFlowy URL, Supabase project ID, and API keys (so they can self-host on their own Hetzner account).
- 2
Implement prompt-caching for Sonnet 4.6: on the second draft request with the same outline, the app re-uses the cached prompt (90% cost reduction on input tokens).
- 3
Add a 'Document Similarity' feature: show the user which other documents in the workspace are similar to the current doc (cosine similarity on Voyage embeddings); allow one-click context insertion.
- 4
Build a 'Pricing Calculator' page: show the user estimated monthly cost based on number of users and AI requests (based on the cost_calculator).
- 5
Add basic version history: every time a section is drafted, store the version in Supabase; user can view/restore previous versions.
Expected output
By Sunday: a working multi-user wiki with AppFlowy backend and an AI sidecar that drafts sections, answers workspace questions, and translates blocks. Cost to run: <$1/mo in API usage (mostly from Sonnet drafts). Enough to demo to 2–3 early customers and collect feedback.
Known gotchas
- !AppFlowy iframe embedding has CORS issues: you'll need to run AppFlowy with 'Access-Control-Allow-Origin: *' flag, or proxy through Supabase Edge Function. DIY version 1 uses iframe + proxy.
- !AppFlowy API is read-only in v2026: you can fetch documents but cannot programmatically create/edit via API. Workaround: store AI-drafted sections in a separate Supabase table, then display side-by-side with AppFlowy doc.
- !Prompt caching in Sonnet has a $0.90/M input token charge (90% discount vs. non-cached). Only worth enabling if you re-use the same system prompt >10 times/day. After 3+ customers, you'll hit this threshold.
- !Voyage embeddings cost $0.06/M tokens. At 100 documents × 2K tokens each = 200K tokens, cost is $0.012 per customer. Cheap, but watch this if scaling to 100+ customers.
- !Lovable doesn't expose AppFlowy's real-time CRDT sync: if two users edit the same doc simultaneously, the AppFlowy iframe handles sync, but the Lovable AI panel won't see live updates. Refresh to see latest state.
- !Multi-user auth per-document is not in the MVP: workspace is single-user. To scale to 5+ internal users per customer, add Supabase RLS and per-document read/write permissions (2–3 weeks of work).
Compliance & risk reality check
A self-hosted document-collaboration platform touches sensitive data (internal playbooks, employee info, financial records, customer lists). If you're building a legal-tech or healthcare wiki, compliance is non-negotiable: HIPAA BAA, GDPR DPA, audit logs, encryption.
GDPR / CCPA for document storage
Documents often contain names, email addresses, and organisational data. If any document is stored on behalf of an EU entity, GDPR requires a Data Processing Agreement (DPA), data localisation (EU server), and retention policies. Hetzner's EU data centres satisfy the residency requirement; you must add a DPA and document your retention policy.
Mitigation: Use Hetzner's EU data centre (Frankfurt region: htz.cloud). Sign DPA with each customer covering AppFlowy + AI processing. Document your purge policy (e.g., 'documents purged 12 months after customer contract ends'). Use Supabase's GDPR features: row-level security (RLS) per-customer, automated data deletion on customer request.
HIPAA BAA for healthcare wikis
If customers use your platform to store clinical notes, patient care protocols, or any protected health information (PHI), HIPAA requires a Business Associate Agreement (BAA), encryption at rest + in transit, audit logging, and access controls. Supabase can support HIPAA with an explicit BAA; AppFlowy self-hosted is your responsibility.
Mitigation: For healthcare customers, mandate HIPAA BAA with Supabase (Pro tier, explicit signing). Use TLS for all connections (Lovable + AppFlowy + Supabase). Implement audit logging: log all document access (user, timestamp, action: view/edit/delete) in an immutable Supabase table. Offer encrypted-at-rest option (AES-256 on Supabase).
SOC 2 Type II for enterprise sales
Fortune 500 customers will ask for SOC 2 Type II attestation. Self-hosting AppFlowy doesn't give you SOC 2 by default—you must document access controls (who can SSH into the VM?), encryption, backup/DR, and immutable audit logs.
Mitigation: Document your current security posture: TLS in transit, AES-256 encryption at rest (Supabase), IP whitelisting for Hetzner, automated daily backups, immutable audit table in Supabase. For full SOC 2 audit, budget $6K–$15K and 3–4 months. Delay until you have 5+ enterprise customers requesting it.
E-discovery and legal hold (for legal tech)
If your platform is used by law firms or in-house legal teams, documents may be subject to litigation holds (cannot delete for duration of lawsuit). You must support legal hold: preventing purge of specific documents or accounts, creating immutable snapshots, and exporting for discovery.
Mitigation: Add a 'Legal Hold' flag per-document: when set, prevent purge and auto-snapshot daily. Create an 'Export for Discovery' feature: user selects a date range and document type, app generates a zip of all docs matching criteria (with metadata: creation date, creator, last-edit date, etc.). Log all holds and exports in audit table.
Build vs buy: the real math
10–14 weeks for an AppFlowy fork with AI integration + compliance setup (GDPR, audit logs).
Custom build time
$13,000–$25,000 (RapidDev standard band). Bump to $25K–$40K if including healthcare compliance (HIPAA BAA setup) or legal-tech features (e-discovery, legal hold).
One-time investment
5–8 months (assuming 12 end-customer accounts at $129/mo each = $1,548/mo recurring, minus $200/mo infrastructure = $1,348/mo contribution; at 70% net margin after customer support, breakeven is $25K ÷ $943/mo = 26.5 months). BUT: with volume discounts on Sonnet (5–10% at >$10K/mo API spend) and higher customer lifetime, breakeven can drop to 6–8 months.
Breakeven vs buying
Buying Notion ($10/user/mo) is the obvious choice for 99% of teams: Notion has $10B in funding, engineers 500+ people, and owns the market. But if you're a knowledge-management consultancy with 5–10 SMB clients pre-committed to a white-label alternative (legal tech, healthcare, SaaS operations), the fork-and-sidecar path wins: you pay RapidDev $13K–$25K upfront, then own 100% of the customer relationship. Reseller margin: if you charge $99–$199/mo and your LLM + infrastructure cost is $0.50/mo, your gross margin is 99%+. At 12 customers × $150/mo = $1,800/mo - $100/mo COGS = $1,700/mo gross profit. The inflection point: once you reach 25+ customers, self-hosting AppFlowy on one beefy Hetzner VM ($100/mo) becomes viable, and your unit economics flatten to nearly 99% margin per customer.
Skip the DIY — RapidDev builds the production version
A Lovable MVP gets you a demo. Production needs auth that doesn't leak data, AI calls that don't bankrupt you, observability when models drift, and code you can audit. That's what we ship.
Discovery call (free)
30 minWe map your exact Document Collaboration Platform with AI use case: who uses it, target volume, AI model choice, integrations, compliance scope. You get a detailed scope document and fixed-price quote within 48 hours.
AI-accelerated build
10–14 weeks for an AppFlowy fork with AI integration + compliance setup (GDPR, audit logs).Our engineers use Claude Code, Lovable, and custom tooling to ship 3–5x faster than agencies. You see weekly progress in a staging environment — not a black box.
Launch + handoff
1 weekWe deploy to your infrastructure, transfer the GitHub repo, set up CI/CD and monitoring, and train your team. You own 100% of the source code, prompts, and model configurations.
What you get
Timeline
10–14 weeks for an AppFlowy fork with AI integration + compliance setup (GDPR, audit logs).
Investment
$13,000–$25,000 (RapidDev standard band). Bump to $25K–$40K if including healthcare compliance (HIPAA BAA setup) or legal-tech features (e-discovery, legal hold).
vs SaaS
ROI in 5–8 months (assuming 12 end-customer accounts at $129/mo each = $1,548/mo recurring, minus $200/mo infrastructure = $1,348/mo contribution; at 70% net margin after customer support, breakeven is $25K ÷ $943/mo = 26.5 months). BUT: with volume discounts on Sonnet (5–10% at >$10K/mo API spend) and higher customer lifetime, breakeven can drop to 6–8 months.
30-min call. Fixed-price quote within 48 hours. No commitment.
Frequently asked questions
How much does it cost to build a white-label doc-collaboration platform?
RapidDev's standard band is $13K–$25K for an AppFlowy fork + AI sidecar (section drafting, Q&A, translation). If you need healthcare compliance (HIPAA BAA), legal-tech features (e-discovery, legal hold), or multi-language support, expect $25K–$40K and 14–18 weeks. DIY with AppFlowy self-host + Lovable sidecar is $25 + API credits (~$50–$100/mo in running cost); it takes 1 weekend but is single-tenant (one instance per customer). Buy-SaaS (Notion at $10/user/mo) is the obvious choice for most teams.
How long does it take to ship a white-label doc-collaboration platform?
RapidDev's fork-and-sidecar build takes 10–14 weeks (AppFlowy customisation + Sonnet integration + audit logging + GDPR setup). A Lovable MVP (AppFlowy iframe + AI panel) runs 1 weekend. Production-grade (multi-user auth per-document, SOC 2 audit-ready, HIPAA support) adds 4–8 weeks.
Can RapidDev build this for my company?
Yes. We've shipped 600+ applications and 80+ AI implementations in production. A white-label doc-collab platform (AppFlowy fork + AI sidecar) is a natural fit—10–14 week timeline, $13K–$25K investment. We recommend starting with a Lovable MVP to test demand with 2–3 early customers, then transitioning to a RapidDev build once you have 5+ committed contracts. Free 30-min consultation: seopartner@rapidevelopers.com.
Why not just buy Notion and resell it?
Notion has $10B valuation and forbids white-labeling—every customer sees 'Powered by Notion' branding. You get 0% gross margin (charge $99/mo, pay $10/user/mo to Notion, lose money at small-customer scale). Reseller discounts are available only for 100+ seats with 2-year commitment. Building a fork-and-sidecar (RapidDev $13K–$25K + AppFlowy OSS) lets you own the UX, control pricing ($99–$199/mo), and achieve 90%+ gross margin.
Is AppFlowy really 'Notion but open source'?
AppFlowy is 80% of Notion's features: block-based editor, databases, full-text search, version history, real-time sync. Missing: AI copilot (you add via Lovable), templates marketplace, integrations (you build custom via API). AppFlowy's UX is 85% of Notion's polish—clean but feels 1–2 years behind. For SMB customers (5–50 users), AppFlowy is plenty; for scaling to 500+ users, you'll want Notion's enterprise infrastructure.
Can I use Gemini 3 Flash for translation instead of Sonnet?
Yes. Gemini 3 Flash is 10x cheaper ($0.075/$0.3 per M tokens vs. $3/$15 for Sonnet) and handles basic translation well (80–85% quality). For casual docs, it's fine. For legal, marketing, or brand-critical docs, Sonnet 4.6 is worth the upgrade (95%+ quality, maintains tone and voice). Implement both: default to Gemini, offer 'Premium translation (Sonnet)' as a paid upgrade.
Do I need pgvector for Q&A, or can I use basic full-text search?
pgvector (semantic search) is better than full-text search for Q&A. Full-text retrieves exact keyword matches; pgvector retrieves semantically similar documents even if keywords don't match. For example, 'What are our values?' retrieves the 'Company Culture' doc (no exact match to 'values') via Voyage embeddings. Cost: $0.06/M tokens for Voyage embeddings; worth it if you have 100+ documents and want high-quality Q&A answers.
What happens if I self-host AppFlowy and the Hetzner VM crashes?
AppFlowy data is stored in PostgreSQL on the Hetzner VM. Set up daily backups to S3 (Backblaze B2: $0.006/GB/mo). If the VM crashes, restore from backup (1–2 hours downtime). For production SLAs (99.9% uptime), move to a managed Postgres service (Supabase Pro $25/mo includes automated backups + PITR = point-in-time recovery).
Can multiple customers share one AppFlowy instance?
Yes, with per-customer workspaces and RLS (row-level security). One Hetzner VM hosts 5–10 customers; one AppFlowy workspace per customer. Cost: $20/mo VM + $25/mo Supabase = $45/mo fixed, split across customers = $4.50/mo per customer. Scales to 100+ customers on a single $100/mo VM with upgrades.
Want the production version?
- Delivered in 10–14 weeks for an AppFlowy fork with AI integration + compliance setup (GDPR, audit logs).
- You own 100% of the code
- AI cost monitoring built in
30-min call. No commitment.