The AI that doesn't just talk. It works. 🐋
Deploys agent swarms, clears your inbox, sends emails, manages your calendar, browses the web, writes code, and runs your life.
All from WhatsApp, Telegram, Discord, or any chat app you already use.
🤖 Multi-Agent Swarms — Fan-out tasks to parallel agents with shared memory — not one bot, an army
🧠 Self-Extensible — Creates its own tools, skills, and automations on the fly
🔒 Runs on Your Machine — Your data never leaves your computer. Private by default
🌐 Website — viralcode.github.io/openwhale
☁️ Managed Hosting — Coming soon
Other AI assistants chat. OpenWhale does. It deploys autonomous agent swarms that fan-out complex work to multiple AI agents running in parallel, coordinate them with shared memory and file locks, and collect results automatically. Connect it to 8 AI providers, let it talk on your behalf through WhatsApp/Telegram/Discord/Twitter/iMessage, browse the web with real browser automation, execute code, manage your calendar, send emails — and basically run your digital life.
An AI workforce, not just an AI assistant. One agent is nice. A coordinated swarm is unstoppable.
git clone https://github.com/viralcode/openwhale.git
cd openwhale
pnpm install
pnpm approve-builds
pnpm run devOpen http://localhost:7777/dashboard → add your API keys → start chatting.
📖 Full Getting Started Guide →
🪟 On Windows? Run the one-line installer in PowerShell — installs everything automatically:
irm https://raw.githubusercontent.com/viralcode/openwhale/main/install.ps1 | iexOr see the Windows Installation Guide → for manual setup.
Switch between any major AI model on the fly. One assistant, unlimited model choices.
| Provider | Top Models |
|---|---|
| Anthropic | Opus 4.5, Sonnet 4.5 |
| OpenAI | GPT-5.2, GPT-5, GPT-4o, o4-mini |
| Google Gemini | Gemini 3 Pro, Gemini 2.5 Pro |
| Qwen | Qwen3-Max, QwQ-Plus, Qwen3-Coder |
| DeepSeek | DeepSeek Chat, Coder, Reasoner |
| Groq | Llama 3.3 70B, Mixtral 8x7B |
| Together AI | Any open-source model |
| Ollama | Local models — no API key! |
Automatic failover between providers. If one fails, the next takes over.
Deploy multiple AI agents in parallel to tackle complex tasks faster. Automatic task decomposition, shared context, advisory file locks, and inter-agent messaging.
"Research quantum computing and write a Python sorting algorithm"
→ Orchestrator fans out to Research Agent + Coder Agent
→ Results synthesized automatically
- Fan-out / Fan-in — Split work across specialized parallel agents
- Shared Context — Agents share data via a namespaced key-value store
- Active Locks — Advisory file locks prevent concurrent write conflicts
- Inter-Agent Messaging — Agents can talk to each other during execution
- A2A Protocol — Google Agent2Agent protocol for interop with LangGraph, CrewAI, AutoGen
Your AI responds across all platforms — unified inbox, one brain.
| Channel | How It Works |
|---|---|
| Scan QR code, uses your personal account | |
| Telegram | Create a bot with @BotFather |
| Discord | Add bot to your server |
| Slack | Slack app integration |
| Twitter/X | Auto-reply to mentions via bird CLI (cookie auth, no API!) |
| iMessage | macOS native — read and send iMessages |
The AI can actually DO things, not just talk about them.
| Category | Tools |
|---|---|
| System | Shell exec, code execution, system info, clipboard, Apple Shortcuts |
| Files | Read/write files, git, zip, codebase search, database queries |
| Documents | PDF creation, PowerPoint slides, Excel spreadsheets |
| Web | Browser automation (Playwright + BrowserOS), web fetch |
| Media | Screenshots, camera, canvas, image analysis, TTS, screen recording, QR codes |
| Communication | Email, iMessage, calendar events |
| Infrastructure | Docker, SSH |
| Memory | Persistent memory, knowledge graphs, multi-step planning |
| Automation | Cron scheduling, self-extensions, skill creator |
| Coordination | Agent listing, session management, shared context, file locks |
Connect to your favorite services — the AI handles the API calls.
| Skill | What It Does |
|---|---|
| GitHub | Manage repos, issues, PRs, commits |
| Notion | Search, create, update pages and databases |
| Google Calendar | View and create events |
| Gmail | Read, send, and search emails |
| Google Drive | Upload, download, list files |
| Google Tasks | Manage your to-do lists |
| Weather | Current conditions and forecasts |
| 1Password | Securely fetch passwords and secrets |
| Apple Notes/Reminders | Native macOS integration |
| Twitter/X | Post, search, timeline, mentions, follow |
| Spotify | Playback control, search, playlists |
| Trello | Boards, lists, and cards |
| ElevenLabs | AI text-to-speech |
| Twilio | SMS and voice calls |
Plus 50+ Markdown Skills from the OpenClaw community.
Two backends to choose from:
| Backend | Best For |
|---|---|
| Playwright | Simple scraping, headless automation, zero setup |
| BrowserOS | Real browser with your extensions, cookies, and logins |
BrowserOS lets the AI control YOUR actual browser — including logged-in sessions, AdBlock, 1Password, and everything else.
The AI can create its own automations. Just ask.
- "Create an extension that checks Bitcoin price every hour and sends it to WhatsApp"
- "Make an extension that reminds me to drink water"
- "Create a daily standup summary extension"
Extensions support cron scheduling, multi-channel output, and persist across restarts.
Remembers everything, even after restarts: long-term facts, daily notes, vector search (local embeddings, no API needed!), and session persistence.
The AI wakes up periodically to check on things — configurations, inboxes, tasks — without you having to ask.
JWT authentication, rate limiting, audit logs, sandboxed code execution, command filtering, multi-user support.
Native SwiftUI menu bar app with quick chat, status monitoring, and channel controls — no browser needed.
| Feature | OpenWhale | ChatGPT | Claude | Typical Chatbot |
|---|---|---|---|---|
| Multi-agent orchestration | ✅ Fan-out/fan-in with shared memory | ❌ No | ❌ No | ❌ No |
| Multi-model support | ✅ 8 providers | ❌ GPT only | ❌ Claude only | ❌ Single model |
| Run shell commands | ✅ Full access | ❌ No | ❌ No | ❌ No |
| Browser automation | ✅ Playwright + BrowserOS | ❌ Limited | ✅ Limited | ❌ No |
| WhatsApp/Telegram/Discord | ✅ All channels | ❌ No | ❌ No | ❌ Maybe one |
| iMessage (macOS) | ✅ Native | ❌ No | ❌ No | ❌ No |
| Twitter integration | ✅ Full API | ❌ No | ❌ No | ❌ No |
| Self-extensible | ✅ Creates own tools | ❌ No | ❌ No | ❌ No |
| Runs locally | ✅ Your machine | ❌ Cloud only | ❌ Cloud only | ❌ Usually cloud |
| Open source | ✅ MIT license | ❌ Closed | ❌ Closed | ❌ Varies |
| Persistent memory | ✅ Vector search | ✅ Limited | ✅ Limited | ❌ Usually no |
| Proactive heartbeat | ✅ Configurable | ❌ No | ❌ No | ❌ No |
| Document | Description |
|---|---|
| Getting Started | Installation, prerequisites, quick start, Docker |
| Providers | AI providers, models, failover configuration |
| Multi-Agent | Fan-out/fan-in, shared context, A2A protocol |
| Channels | WhatsApp, Telegram, Discord, Slack, Twitter, iMessage |
| Tools | Complete reference for all 40+ built-in tools |
| Skills | API skills and Markdown community skills |
| Browser Automation | Playwright vs BrowserOS |
| Memory | Long-term memory, vector search, session persistence |
| Heartbeat | Proactive agent configuration |
| Extensions | Self-extension system and scheduling |
| Dashboard | Web dashboard and setup wizard |
| macOS App | Native menu bar app |
| API | REST API reference and OpenAI-compatible endpoints |
| Security | Authentication, rate limiting, audit, sandboxing |
| Windows Setup | Installing & running on Windows 10/11 |
| Configuration | Full .env reference |
| CLI | CLI commands and npm scripts |
| Troubleshooting | Common issues and fixes |
src/
├── agents/ # Multi-agent orchestration, inter-agent comms, A2A protocol
├── auth/ # JWT, API keys, sessions
├── channels/ # WhatsApp, Telegram, Discord, Slack, Twitter, iMessage adapters
├── cli.ts # Interactive terminal interface
├── daemon/ # Background service (launchd on macOS)
├── dashboard/ # Web admin panel + coordination dashboard
├── db/ # SQLite/PostgreSQL with Drizzle ORM
├── gateway/ # Hono-based HTTP API routes
├── heartbeat/ # Proactive agent scheduler
├── integrations/# Google APIs (Calendar, Gmail, Drive, Tasks)
├── memory/ # Persistent memory, embeddings, vector search
├── providers/ # Anthropic, OpenAI, Google, Groq, Qwen, DeepSeek, Together, Ollama
├── security/ # Rate limiting, audit logs, sandboxing, command filtering
├── sessions/ # Persistent conversations, fan-out auto-detection, compaction
├── skills/ # GitHub, Notion, Spotify, Weather, Apple, Twitter, etc.
├── tools/ # 40+ tools — files, browser, code exec, screenshots, coordination
└── voice/ # Voice interaction support
PRs welcome! If you find a bug or have an idea, open an issue. Keep it friendly.
MIT — do whatever you want with it.
Built with ❤️ by Jijo John