The open source AI sales development platform. AI cold calling software, lead generation, sales automation, and a full AI SDR team — all running locally on your desktop. No SaaS fees. No cloud dependency. No data ever leaves your machine.
🎬 Watch the Launch Video · 🚀 Quick Start · 📸 Screenshots · ✨ Features · 🏗️ Architecture · 📦 Download · 🤝 Contributing
| Dashboard | Pipeline Kanban | War Room (Live AI Call) |
|---|---|---|
 |
 |
 |
| KPI dashboard with pipeline analytics | Kanban CRM with AI execution tracking | Live AI cold call with sentiment & coaching |
| AI Persona Builder | Lead Hunter | Call Intelligence |
|---|---|---|
 |
 |
 |
| Voice engine architect with emotion matrix | AI lead generation with ICP targeting | Call analytics with outcome tracking |
| Lead Detail | Sales Coach | Settings |
|---|---|---|
 |
 |
 |
| 360° lead view with timeline & notes | Objection sparring with 3 difficulty levels | Voice engine, phone line & API key config |
OpenCloser is an open-source AI sales development platform that runs entirely on your desktop. It gives you an autonomous AI sales team: a strategist to build your ICP, a researcher to hunt leads, a voice caller to dial prospects, a coach to train your rebuttals, and a manager to analyze every call — all running locally.
Why open source AI SDR? Traditional SDR platforms charge $280–$1,000+/month per seat. OpenCloser is free, open source, and self-hosted — your data stays on your machine, your AI keys stay in your browser, and you own the entire stack.
| Traditional SDR Platform | OpenCloser | |
|---|---|---|
| Strategy | Hire a consultant ($5K+/mo) | AI Strategist generates ICP using SPIN & Challenger in minutes |
| Research | SDR manually Googles leads (hours) | AI Researcher hunts, scores & qualifies leads automatically |
| Cold Calling | SDR dials 50 calls/day, burns out | AI Caller dials with perfect pitch, never tired |
| Coaching | Manager reviews recordings (hours) | AI Coach gives instant post-call analysis |
| Training | Roleplay sessions (awkward, infrequent) | AI Sparring Partner available 24/7, adjustable difficulty |
| Analytics | Spreadsheets and gut feelings | Real-time dashboards, sentiment analysis, pipeline tracking |
| Cost | $280–$1,000+/mo per seat | Free. Open source. MIT license. |
| Data Privacy | Cloud-hosted, vendor-controlled | Local-first. Zero cloud dependency. |
- Solo founders who need to sell but can't afford a sales team
- Early-stage startups that want enterprise-level sales ops from day one
- Sales teams looking to augment human reps with AI intelligence
- Developers who want to build on top of an open sales AI framework
- Privacy-conscious teams who can't send customer data to SaaS platforms
Generates your Ideal Customer Profile (ICP) using SPIN & Challenger frameworks. Asks the right questions during onboarding to deeply understand your market. Builds targeted outreach strategies automatically.
Autonomous lead hunting with scoring and qualification. Full local-first CRM with Kanban pipeline management. Industry-aware demo mode with 5 keyword categories when no API key is configured.
Real-time AI voice agent powered by Google Gemini (also supports OpenAI Realtime and ElevenLabs ConvAI). Built-in Twilio phone lines — the AI speaks directly into live calls from the War Room (mock mode until you add credentials; bundled cloudflared tunnel for the audio stream). Virtual audio bridge for VoIP apps. Live transcription, sentiment analysis, and objection detection. Power dialing mode for high-volume outreach.
Objection sparring trainer with 3 difficulty levels (Rookie, Pro, Elite). Practice handling 12 objection archetypes against an AI prospect. Real-time encouragement and post-session scoring with specific improvement tips.
Post-call AI debrief with sentiment analysis and key insights. Auto-generated follow-up emails. Call analytics dashboard with conversion metrics. Pipeline intelligence with live Kanban board.
- Node.js (v20+)
- Rust (latest stable)
- Tauri prerequisites (system dependencies per platform)
git clone https://github.com/issacops/opencloser.git
cd opencloser
npm install
npm run tauri devOpenCloser works fully in demo mode with no API key — 10 realistic seed leads, simulated AI calls, and all features functioning with fallback data. Perfect for evaluation.
Add any of these in Settings → Voice Engine:
| Provider | Key |
|---|---|
| Google Gemini | GEMINI_API_KEY for onboarding, lead hunting, call analysis |
| OpenAI Realtime | openai_api_key for voice calling through the built-in relay |
| ElevenLabs ConvAI | elevenlabs_api_key + Agent ID for the most human-like voice |
| Local Ollama | none — fully offline AI (Settings → Offline AI); takes priority over cloud when configured |
Then connect a Twilio phone line in Settings → Phone — or skip credentials entirely and phone-line calls run in mock mode.
opencloser/
├── src/ # React 19 frontend (TypeScript strict mode)
│ ├── features/
│ │ ├── crm/ # Kanban pipeline, Dashboard, Settings, Persona
│ │ ├── voice/ # War Room, Post-Call Debrief, Objection Trainer
│ │ ├── hunter/ # Lead hunting engine
│ │ └── onboarding/ # AI onboarding + ICP generation
│ ├── stores/ # 5 Zustand stores (call, navigation, persona, ...)
│ ├── services/ # Typed Tauri invoke wrappers (leads, AI, Twilio)
│ ├── test/ # 74 Vitest tests across 8 suites
│ └── components/ # AppShell, shared UI
├── src-tauri/ # Rust backend
│ └── src/
│ ├── ai/gemini.rs # Gemini API + industry-aware demo mocks
│ ├── ai/ollama.rs # Offline Ollama path (priority over cloud)
│ ├── twilio/ # Twilio REST + Media Streams + cloudflared tunnel
│ ├── db/ # SQLite schema + 10 seed leads + migrations
│ ├── relay/ # WebSocket proxy for OpenAI/ElevenLabs
│ └── lib.rs # 24 Tauri commands registered
└── public/
└── audio-processor.worklet.js # Zero-latency PCM capture
| Layer | Technology |
|---|---|
| Frontend | React 19, TypeScript strict, Tailwind CSS 3, Vite 6, Zustand 5 |
| Desktop Runtime | Tauri 2.10 |
| Backend | Rust (rusqlite, reqwest, tokio, tokio-tungstenite) |
| Database | SQLite — local-first, zero cloud dependency |
| AI | Google Gemini 2.5 Flash, OpenAI Realtime, ElevenLabs ConvAI, local Ollama |
| Voice | Web Audio API + AudioWorklet (zero-latency PCM capture), Twilio Media Streams |
| Telephony | Twilio (mock mode included), bundled cloudflared quick tunnel |
| Tests | Vitest + React Testing Library (74 tests, 8 suites) + 25 Rust tests |
| CI/CD | GitHub Actions — lint, format, knip, test, build, multi-platform release |
| Platform | Architecture | Download |
|---|---|---|
| macOS | Apple Silicon (ARM64) | .dmg |
| macOS | Intel (x64) | .dmg |
| Windows | x86_64 | .msi / .exe |
| Linux | x86_64 | .deb / .AppImage / .rpm |
All packages available on the Releases page.
| Metric | |
|---|---|
| TypeScript | Strict mode, 0 errors |
| Tests | 74 JS + 25 Rust passing |
| Codebase | ~14,400 lines (TS: 10,749 + Rust: 2,869 + CSS: 816) |
| Bundle | 404KB main + 9 lazy-loaded chunks |
| Demo mode | No API key required — full fallback data, mock phone lines |
- All data stays local — SQLite database on your machine, never synced to any server
- API keys stored in your browser's localStorage — never transmitted except to the AI provider you choose; they are sent as headers, never in URLs
- Offline capable — configure Ollama (Settings → Offline AI) and interviews, research, call analysis and objection training run entirely local
- Open source — every line auditable. MIT license.
Option A — Built-in Twilio phone line (recommended): add your Twilio credentials in Settings → Phone (or stay in mock mode), toggle auto-tunnel, then press PHONE LINE in the War Room header. The AI speaks directly into the live call — no virtual cables required.
Option B — Virtual audio cable (VoIP apps like Phone Link / FaceTime):
The built-in Audio Setup Wizard guides you through configuration, or skip entirely for VoIP-only mode.
We welcome contributions! See AGENTS.md for the full developer guide.
# Setup
npm install
# Quality gates (run before committing)
npm run lint # ESLint + strict tsc
npm run format:check
npm run knip # no unused exports
npm test # Vitest suites
# Rust gates (in src-tauri/)
cargo fmt --check
cargo clippy --all-targets -- -D warnings
cargo test
# Run desktop app
npm run tauri dev- Fork the repository
- Create a feature branch (
git checkout -b feat/amazing-feature) - Commit your changes with clear messages
- Ensure
npm run lintandnpm testpass - Open a Pull Request
MIT License — see LICENSE for details.
- Tauri — Making native desktop apps accessible to web developers
- Google Gemini — The AI backbone
- OpenAI — Realtime voice API
- ElevenLabs — Most human-sounding voice synthesis
- Lucide Icons — Beautiful icon set
Built with ❤️ for sales teams who want an unfair advantage.
⭐ Star This Repo · 🐛 Report Bug · 💡 Request Feature · 📖 Developer Guide · 📦 Releases
Keywords: open source AI SDR AI sales platform AI cold calling software self-hosted AI sales agent sales automation AI voice caller desktop CRM lead generation open source CRM Tauri React Rust TypeScript Gemini AI sentiment analysis sales pipeline objection handling Twilio Ollama local AI