The Ultimate, Locally-Hosted, Zero-Ban AI Content Operating System for LinkedIn.
Write like a human. Scale like a machine. No SaaS subscriptions. No vendor lock-in.
English | ΰ€Ήΰ€Ώΰ€ΰ€¦ΰ₯ β /hi/dashboard live via next-intl
| $0/mo vs $199 Taplio Pro |
n8n Node Builder 12 Prebuilt Automation Flows |
Real Webhook Schedule Zapier / Make / Buffer Dispatch |
54+ AI Providers Frontier Β· Fast Β· Local Ollama |
Telegram Remote 24/7 Phone Control via Bot |
Zero-Ban No unofficial scrapers, 100% Safe |
LUNVO is designed, built, and open-sourced by THE Ξ LAB (The Pi Lab).
- Company LinkedIn: THE Ξ LAB on LinkedIn
- Live Production URL: https://lunvo-tawny.vercel.app/
- Official Website: https://www.thepilab.in
- Email Contact:
hello@thepilab.in
Most LinkedIn AI tools are $39-$199/mo closed SaaS wrappers around a single GPT-3.5/4 prompt β plus they use unauthorized browser extensions that can get your LinkedIn account permanently restricted or banned.
| Feature | Taplio | Supergrow | AuthoredUp | LUNVO 2.0 (Open Source) |
|---|---|---|---|---|
| Price | $39-$199/mo | $19-$69/mo | $19/mo | $0 β MIT Open Source |
| Node Automation | β None | β None | β None | β n8n-Style 2D Visual Canvas (12 Prebuilts) |
| Real Auto-Scheduling | Risky Scraper | Risky Scraper | Manual only | β Outbound Webhooks (Zapier / Make / Buffer / Pipedream) |
| Remote Phone Control | β None | β None | β None | β
24/7 Telegram Companion Bot (/idea, /approve, /schedule) |
| AI Provider Choice | Locked to OpenAI | Locked | None | β 54+ Providers: Claude, Gemini, DeepSeek, Groq, Mistral, Ollama |
| Voice Cloning | Generic prompt | Static template | Manual | β Voice DNA (4 sliders + auto-extract from past posts) |
| Autonomous Pipeline | Single prompt | Single prompt | None | β 3-Agent Neural Pipeline: Scout β Writer β Critic |
| Viral Score Predictor | Subjective / Basic | No | No | β Data-Backed Analytics Predictor (Trained on Real Posts) |
| Carousel PDF Studio | Pro only ($99+) | No | No | β 5 Executive Templates, 1080Γ1350 High-Reach PDF Export |
| Anti-AI Humanizer | No | No | No | β Banned ClichΓ© Stripper + Burstiness Variation (90%+ Human) |
| Omni Repurposer | No | No | Basic | β 1-Click: Post or YouTube Transcript β Thread + Newsletter |
| AI Image Studio | Basic / DALL-E only | No | No | β 6 Pro Styles + Pollinations Neural Visuals + Canvas Cards |
| Ban Risk | Medium (unofficial API) | Medium | Low | β Zero β Official Webhooks & Native System Clipboard Only |
| Self-Hostable | No | No | No | β 1-Click Vercel / Railway / Docker in 30 seconds |
| IDE MCP Server | No | No | No | β
Yes, Claude Desktop & Cursor ready (npx lunvo-mcp) |
LUNVO gives you the power of Taplio + Supergrow + ViralBrain + AuthoredUp β completely free, private, and account-safe.
Live Web Preview β https://lunvo-tawny.vercel.app/ Watch LUNVO 2.0 execute the Autonomous 3-Agent Neural Pipeline and n8n-Style Workflow Canvas live in action.
π Autoplaying Video Tour (loop): Scout extracts 3 viral hooks β Writer streams draft word-by-word with Voice DNA β Critic scores 92/100 β Visual node automation executes the pipeline.
π¬ Full 2m 48s 1080p HD video with audio dekhne ke liye upar Watch Full 1080p Video badge ya video image par click karein!
πΈ UI Showcase & Feature Gallery β Direct From Live Production (Click to toggle)
The entry door to the LinkedIn OS β featuring real-time interactive previews and zero-friction access.
Consolidated view of your LinkedIn growth streak, quick actions, pipeline status, and content health.
Commission posts via the 3-agent neural pipeline (Scout, Writer, Critic) with active Voice DNA injection and built-in Image Studio supporting 6 professional visual styles.
Instant audit of hook impact, mobile readability, dwell-time estimation, and Anti-AI Humanizer score.
Full interactive pan, zoom (30% to 200%), node wiring, and live topological pipeline execution.
12 prebuilt automation recipes β Daily RSS News to Post, YouTube Transcript Repurposer, Bullet Notes to Carousel PDF, Quality Gatekeeper, and more.
Schedule posts directly to Zapier, Make, and Buffer with zero account risk (no scrapers, no cookie hijacking).
100% BYOK supporting 50+ providers (Frontier, Aggregators, Fast Inference, Open & Cheap, Privacy, Gateways, Local & Free) with multiple keys and automatic failover.
Remote command interface for smartphone control β generate ideas via /idea, rewrite via /humanize, and /approve directly to queue.
Direct native IDE integration for Claude Desktop, Cursor, and Antigravity with 8 registered tool endpoints (npx lunvo-mcp).
graph TD
User([User / Browser]) --> Shell[AppShell + Studio UI]
Shell --> Vault[BYOK Profile Vault<br/>54+ Providers Β· Client-Side Storage]
subgraph "Neural Generation Engine"
Orchestrator[3-Agent Orchestrator] --> Scout[π΅οΈ Scout Agent<br/>Angle + 3 Hooks]
Scout --> Writer[βοΈ Writer Agent<br/>Voice DNA Injected]
Writer --> Critic[π§ Critic Agent<br/>Virality & Dwell Time]
Critic --> Humanizer[π‘οΈ Humanizer Engine<br/>Banned ClichΓ© Stripper & Burstiness]
end
Shell --> Orchestrator
Vault --> Orchestrator
subgraph "Automation & Scheduling"
Workflow[Visual Canvas 2D<br/>workflowRunner.ts] --> Nodes[12 Prebuilt Node Flows]
Nodes --> Dispatcher[webhookDispatcher.ts<br/>SSRF-Guarded Outbound Queue]
Dispatcher --> Outbound[Zapier / Make / Buffer / Pipedream]
TelegramBot[botService.ts<br/>24/7 Telegram Companion] --> Orchestrator
TelegramBot --> Dispatcher
end
subgraph "Headless Tools"
MCPServer[mcp/server.ts<br/>npx lunvo-mcp] --> Orchestrator
MCPServer --> Dispatcher
end
Single source of truth supporting all AI providers with zero adapter bloat:
- Frontier: OpenAI, Anthropic, Google Gemini, xAI Grok, DeepSeek, Mistral.
- Aggregators: OpenRouter, GitHub Models, HuggingFace Router.
- Fast Inference: Groq, Cerebras, SambaNova, NVIDIA NIM, Fireworks AI.
- Open & Affordable: Together AI, DeepInfra, Hyperbolic, Novita AI, Nebius, Lambda AI, Scaleway, Baseten, Crusoe, IO.NET, Replicate, MiniMax, Moonshot Kimi, Zhipu GLM, Qwen (Alibaba Cloud), Upstage Solar, Sarvam AI, Clarifai, Chutes, DigitalOcean Gradient.
- Privacy First: Venice AI, Perplexity Sonar, Cohere.
- Gateways: Vercel AI Gateway, Cloudflare Workers AI, AIHubMix, 302.AI, LangDB, Eden AI, Dify, Letta, SAP AI Core.
- Local & Offline: Ollama, LM Studio, llama.cpp, Jan, Ollama Cloud.
- Auto-Failover: Automatically handles rate limits and non-retryable provider errors.
- Drag-and-drop node graph with canvas zoom (40%-180%), infinite placement, and 1-key fit view ('F').
- Smooth n8n-Style S-Curves: Organic cubic bezier connection wires with minimum curvature offsets and real-time interactive preview wire drawing on an unclipped SVG canvas.
- 12 Production Templates: RSS News to Post, YouTube Transcript Repurposer, Bullet Notes to Carousel PDF, Quality Gatekeeper (Retry if Score < 85), Multi-Platform Matrix, and more.
- Full import/export of pipeline configurations as portable JSON.
- 4 Parametric Sliders: Formality, Humor, Directness, Punchiness.
- Ingest past successful posts to automatically extract and train your custom Voice DNA.
- Local-first storage guarantees privacy β your voice profile never leaves your browser unless you configure an external connector.
- Scans and eliminates corporate buzzwords and AI clichΓ©s (
delve,tapestry,landscape,game-changer,synergy,elevate,in conclusion, etc.). - Calculates Burstiness (sentence length variance) and ensures natural conversational rhythm.
- Real-time
isHumanScoremeter targeting 90%+ human readability.
- Calibrated on real LinkedIn post engagement analytics.
- Evaluates Content Length (mobile sweet spot 250-400 chars), Hashtag Count (optimal 4-6), Post Timing (peak hours & days), Hook Impact (vulnerability, contrarian, data-backed), and CTA Specificity.
- Generates pixel-perfect 1080Γ1350 documents optimized for LinkedIn document carousels.
- 5 Executive Themes: Minimal, Bold Corporate, Dark Mode Neon, Clean Editorial, and Card Deck.
- Instant client-side PDF export using
jspdf.
- Input any raw note, blog article, or YouTube URL.
- Automatically fetches transcripts via caption tracks (zero Google API keys needed).
- One click converts source material into a LinkedIn Post, a Twitter/X Thread, and an Email Newsletter.
- Completely eliminates LinkedIn account ban risk by avoiding unofficial scrapers and browser automation.
- Scheduled dispatch directly to your Zapier, Make.com, Buffer, or Pipedream webhook.
- Enterprise-Grade SSRF Protection: Validates URLs against private IP subnets, loopback, cloud metadata endpoints, octal/hex encodings, userinfo smuggling, and validates HTTP 3xx redirect targets.
- Control LUNVO from anywhere on your mobile phone.
/idea <topic>β Generates contrarian angles and complete drafts./approveβ Pushes the draft directly to your publishing queue./humanizeβ Rewrites with the Anti-AI engine./trendingβ Summarizes real-time tech news from RSS feeds.- Built-in memory safety with automatic cache pruning.
- Plug LUNVO directly into Claude Desktop, Cursor, or Antigravity IDE:
npx lunvo-mcp
- 8 registered MCP tools:
search_trending,analyze_draft,generate_pipeline_post,humanize_post,train_voice_dna,repurpose_content,execute_workflow, andschedule_post.
Visit the live deployment at https://lunvo-tawny.vercel.app/.
# 1. Clone the repository
git clone https://github.com/the-pi-lab/Lunvo.git
cd Lunvo
# 2. Install dependencies
npm install
# 3. Configure environment variables (BYOK)
cp .env.example .env.local
# 4. Start development server
npm run dev
# -> Navigate to http://localhost:3000# Build and run containerized LUNVO
docker compose -f docker-compose.prod.yml up -d --build
# -> Accessible on http://localhost:3000βοΈ Environment Variables Reference β Click to expand
| Variable | Required | Description |
|---|---|---|
NEXT_PUBLIC_APP_URL |
No | Public URL (https://lunvo-tawny.vercel.app/ or http://localhost:3000) |
GEMINI_API_KEY |
Optional | Free platform inference (aistudio.google.com) |
GROQ_API_KEY |
Optional | Ultra-fast inference (console.groq.com) |
OPENAI_API_KEY |
Optional | OpenAI API key |
ANTHROPIC_API_KEY |
Optional | Anthropic Claude API key |
TELEGRAM_BOT_TOKEN |
Optional | Bot token from @BotFather for Telegram Companion |
TELEGRAM_CHAT_ID |
Optional | Authorized Telegram Chat ID for remote commands |
TELEGRAM_WEBHOOK_SECRET |
Optional | Secret token for securing /api/telegram webhook |
CRON_SECRET |
Optional | Bearer secret for /api/cron/rss-refresh endpoint |
UPSTASH_REDIS_REST_URL |
Optional | Distributed rate limiter storage for production clusters |
UPSTASH_REDIS_REST_TOKEN |
Optional | Distributed rate limiter auth token |
LUNVO is created and maintained by THE Ξ LAB:
- LinkedIn Page: https://www.linkedin.com/company/the-%CF%80-lab/
- Website: https://www.thepilab.in
- Live Demo: https://lunvo-tawny.vercel.app/
- Founder & Architect: Vinayak Mahavar
If LUNVO saves you the $199/month cost of closed SaaS tools, please star this repository on GitHub!
Share the love: Tweet I just found LUNVO β open source Taplio killer. BYOK. Zero ban. Self-host in 30s. github.com/the-pi-lab/Lunvo π
MIT License Β© THE Ξ LAB β MAHAVAR VINAYAK DILIPKUMAR
Account Safety Guarantee: LUNVO never interacts with unauthorized, private, or reverse-engineered LinkedIn APIs. Publishing is strictly performed via standard system clipboard or verified outbound webhook dispatchers to your chosen scheduler (Zapier, Make, Buffer). Your LinkedIn credentials, cookies, and account remain 100% safe.
