AI-powered LinkedIn post creation, scheduling, and publishing — as a desktop app.
From idea to published LinkedIn post in minutes, not hours.
Afterposten is a local-first desktop app that turns rough ideas into polished LinkedIn posts using AI. Write an idea, generate multiple draft variants, pick your favorite, attach an AI-generated image, schedule it, and publish to LinkedIn via n8n webhooks — all from one clean interface.
No cloud accounts. No subscriptions. Your data stays on your machine.
Create → Generate → Edit → Schedule → Publish
| Feature | Description |
|---|---|
| 💡 Idea → Draft | Paste an idea, AI generates a full LinkedIn post with hook, body, CTA, and hashtags |
| 🔄 Variant Generation | Get 3–5 different angles on the same idea, pick the best one |
| 🖼️ AI Image Generation | Generate post images with style presets (clean-tech, editorial, bold-minimal, etc.) |
| #️⃣ Smart Hashtags | AI suggests 5–15 relevant hashtags based on your content |
| 🎙️ Voice Input (STT) | Speech-to-text via Whisper — speak in any language, get English text for ideas and knowledge base |
| 📅 Scheduling | Set a date/time in your timezone, the scheduler handles the rest |
| 🔌 n8n Integration | Publish via webhook with multipart/form-data — text, image, metadata |
| 🔒 Encrypted Credentials | Publisher auth tokens encrypted at rest (AES-256-GCM) |
| 🧠 Knowledge Base (RAG) | Embedded Vectra vector DB for voice/style context — add URLs, text docs, or let it auto-learn from your posts |
| ⚡ Real-time Streaming | Watch the AI draft appear word-by-word as it generates |
| 🖥️ Desktop App | Runs as a native Electron app on macOS, Windows, and Linux |
┌─────────────────────────────────────────────────────┐
│ Electron Shell │
│ ┌───────────────────────────────────────────────┐ │
│ │ Next.js (App Router) │ │
│ │ ┌─────────┐ ┌──────────┐ ┌─────────────┐ │ │
│ │ │ Posts │ │ Settings │ │ Post Editor │ │ │
│ │ │ List │ │ Page │ │ + AI Panel │ │ │
│ │ └────┬────┘ └────┬─────┘ └──────┬──────┘ │ │
│ │ │ │ │ │ │
│ │ └─────────────┼───────────────┘ │ │
│ │ ▼ │ │
│ │ API Routes (13) │ │
│ │ ┌─────┬──────┬───────┬──────┐ │ │
│ │ │ AI │ Pub │ Sched │ CRUD │ │ │
│ │ └──┬──┴──┬───┴───┬───┴──┬───┘ │ │
│ │ │ │ │ │ │ │
│ │ ▼ ▼ ▼ ▼ │ │
│ │ OpenAI n8n Poller SQLite Vectra │ │
│ │ API Webhook (Prisma) (RAG) │ │
│ └───────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────┘
Grab the latest release for your platform from Releases:
| Platform | Format |
|---|---|
| macOS (Apple Silicon) | .dmg / .zip |
| macOS (Intel) | .dmg / .zip |
| Windows | .exe (NSIS installer) / .zip |
| Linux | .AppImage / .deb |
# Clone the repo
git clone https://github.com/scrobot/afterposten.git
cd afterposten
# Install dependencies
pnpm install
# Set up the database
pnpm db:migrate
# Configure environment
cp .env.example .env
# Edit .env with your OpenAI API key and encryption keyCreate a .env file in the project root:
# Required: Your OpenAI API key
OPENAI_API_KEY=sk-...
# Required: 64-char hex string for encrypting publisher credentials
# Generate one: node -e "console.log(require('crypto').randomBytes(32).toString('hex'))"
ENCRYPTION_KEY=your-64-char-hex-string
# Optional: Database location (defaults to file:./dev.db)
DATABASE_URL=file:./dev.db# Start Next.js dev server (web mode)
pnpm dev
# Start in Electron (desktop mode)
pnpm dev:electron
# Run unit tests (37 tests)
pnpm test
# Type check
pnpm typecheck
# Build Electron app for current platform
pnpm build:electronAfterposten publishes via n8n webhooks. Set up a workflow in n8n that:
- Receives a Webhook trigger (POST, multipart/form-data)
- Reads the fields:
text,hashtags(JSON array),postId,scheduledAt - Reads the binary image from the
mediaFilefield (configurable) - Posts to LinkedIn using the LinkedIn node
| Mode | Description |
|---|---|
| None | No auth (for local n8n instances) |
| Custom Header | Send a custom header (e.g., X-Auth-Key: secret) |
| Bearer Token | Standard Authorization: Bearer <token> |
All auth credentials are AES-256-GCM encrypted at rest in the local database.
pnpm test # 37 unit tests (Vitest)
pnpm test:watch # Watch mode
pnpm test:e2e # E2E tests (Playwright, requires browser install)
pnpm typecheck # TypeScript strict mode| Suite | Tests | What it covers |
|---|---|---|
schemas.test.ts |
19 | Zod schema validation for all AI contracts |
tz-utils.test.ts |
10 | Timezone conversion (CET/CEST, midnight, boundaries) |
publisher-adapter.test.ts |
5 | n8n adapter auth modes, FormData, sanitized logging |
crypto.test.ts |
3 | AES-256-GCM encrypt/decrypt, random IV |
| Layer | Technology |
|---|---|
| Desktop | Electron 40 |
| Frontend | Next.js 16 (App Router), React 19, TypeScript |
| AI | Vercel AI SDK (streamObject), OpenAI GPT-4o / GPT-Image-1 |
| RAG | Vectra (embedded vector DB, file-backed, text-embedding-3-small) |
| Database | SQLite (Prisma 7 + better-sqlite3 driver adapter) |
| Scheduling | Polling scheduler with atomic locking and exponential backoff |
| Publishing | n8n webhooks via multipart/form-data |
| Security | AES-256-GCM encryption for auth credentials |
| Testing | Vitest + Playwright |
MIT © scrobot