The AI Assistant That Actually Does Things
In January 2026, an open-source project took Silicon Valley by storm: ClawdBot (now Moltbot). Unlike traditional AI assistants that merely chat, Moltbot actually takes action: it manages your emails, organizes your calendar, checks you in for flights, controls your smart home devices, and executes commands on your computer.
Created by Peter Steinberger, an Austrian developer and founder of PSPDFKit, Moltbot was born from a personal need. After three years away from technology, Steinberger rediscovered his passion by building an AI assistant that could genuinely manage his digital life. The result? A tool that garnered over 60,000 GitHub stars in just weeks and sent Cloudflare’s stock surging 14% in a single day.
60K+
GitHub Stars
8,900+
Discord Members
14%
Cloudflare Surge
What Exactly Is Moltbot?
Moltbot is a self-hosted personal AI assistant that runs on your own hardware (Mac Mini, Linux server, Raspberry Pi, or VPS). Instead of living in a browser tab, it integrates directly with the messaging apps you already use:
📱 Multi-Platform
WhatsApp, Telegram, Discord, Slack, Signal, iMessage, Matrix, Teams, and more
🔒 Private by Default
All your data stays on your machine – nothing goes to the cloud
🤖 Proactive
Can reach out with reminders, updates, and insights without you asking
🔧 Extensible
Plugin-based architecture with a growing skills marketplace
Moltbot’s true power lies in its ability to maintain persistent context and execute complex multi-step tasks. It acts like “Claude with hands” – combining the power of a large language model with actual action capabilities.
Key Features and Use Cases
🎯 Daily Automation
- Email management: Automatic unsubscribing, newsletter archiving, urgent message prioritization, professional response drafting
- Smart calendar: Meeting scheduling, conflict resolution, contextual reminders
- File management: Automatic folder organization, document processing, Notion synchronization
- Smart home: Philips Hue light control, IoT device management
💻 Development and Productivity
- Shell command and script execution
- GitHub project management (PR creation, debugging)
- Automated compilation and testing
- Remote code deployment from a phone
🔔 Proactive Monitoring
Thanks to its “heartbeat” engine and cron integration, Moltbot can monitor conditions and react automatically:
- Server monitoring and downtime alerts
- Stock price tracking with custom thresholds
- Log file monitoring
- Regular system health checks
💡 Advanced Use Case Example: One user configured Moltbot to monitor their work emails, automatically create tasks in Todoist for important requests, update their Google Calendar accordingly, and send them a morning voice summary via ElevenLabs. All without manual intervention!
How Does It Work?
Moltbot operates via a lightweight local gateway that handles communication, tools, and persistent state. This gateway runs in the background as a system service, processing incoming messages while orchestrating LLM-driven responses and actions.
Technical Architecture
- Persistent sessions: Long-term memory stored locally, enabling context retention across days, weeks, or months
- Tool integration layer: Modular “skills” written in TypeScript/JavaScript providing controlled access to system resources
- Proactive engine: Cron jobs, webhooks, and heartbeat checks enable autonomous behavior
- Multi-channel bridge: Unified interface connecting messaging platforms to the central agent
- Sandbox options: Docker integration and permission controls for secure execution
Quick Installation
# One-line installation on macOS or Linux curl -fsSL https://clawd.bot/install.sh | bash # Then run the setup wizard moltbot onboard –install-daemon # Start the gateway moltbot gateway –port 18789 –verbose # Send a test message moltbot message send –to +1234567890 –message “Hello from Moltbot”
⚠️ Important Security Considerations
🔐 SECURITY WARNING: Moltbot’s power comes with significant risks. “Actually doing things” means “can execute arbitrary commands on your computer.”
Security experts have raised several legitimate concerns:
Main Risks
- Prompt injection: A malicious WhatsApp message could potentially trick Moltbot into executing unwanted actions
- Root access: Moltbot requires elevated permissions for certain operations, which is both powerful and dangerous
- Exposed instances: Shodan scans have revealed hundreds of misconfigured instances exposed on the web
- Supply chain vulnerabilities: The skills marketplace (ClawdHub) could be exploited to distribute malicious code
Security Best Practices
- DO NOT install on your main machine with your SSH keys, API credentials, and password manager
- Use a dedicated VPS or separate Mac Mini with throwaway accounts
- Enable sandbox mode for group sessions and non-primary channels:agents.defaults.sandbox.mode: “non-main”
- Configure strict whitelists for who can interact with the assistant
- Review all skills before installation from the marketplace
- Keep Moltbot updated with the latest security patches
Creator’s Note: Peter Steinberger himself describes running Moltbot on a main machine as “spicy” (risky). The security/utility trade-off is real and requires thoughtful configuration.
The Ecosystem and Community
The project has created a vibrant community movement:
- 50+ active contributors on the GitHub repository
- 8,900+ member Discord sharing configurations and use cases
- ClawdHub: skills marketplace with hundreds of integrations
- “Vibe Coding”: trend where developers submit PRs largely written by AI
The community has developed integrations with dozens of services:
🏢 Productivity
Gmail, Google Calendar, Todoist, Obsidian, GitHub, WordPress
💪 Health & Fitness
WHOOP, fitness tracking, personalized wellness routines
🎵 Entertainment
Spotify, playlist management, content recommendations
🛠️ Development
Code execution, PR creation, multi-machine debugging
Industry Impact
Moltbot’s meteoric rise has had measurable effects:
- Mac Mini M4 stock shortage: Thousands of users wanting to host Moltbot created a global shortage
- Cloudflare stock +14%: Enthusiasm around Moltbot propelled Cloudflare’s stock price, as its infrastructure is used to run the assistant
- Crypto scams: Fake “ClawdBot” tokens reached $16M market cap before collapsing 90% – the creator had to publicly warn he never launched any crypto
- AI agent debate: Palo Alto Networks cites AI agents as “the biggest insider threat of 2026”
Key TechCrunch Quote: “By building a tool to solve his own problem, Steinberger showed the developer community what AI agents could actually accomplish, and how autonomous AI might finally become genuinely useful rather than just impressive.”
Models and Configuration
Moltbot supports multiple model backends:
- Claude (Anthropic): Recommended for best performance (via API)
- Llama 3.3 70B: Powerful open-source option
- Minimax M2.1: Popular alternative in the community
- Docker Model Runner (DMR): For running LLMs locally without API fees
Recommended Configuration
{ “model”: “claude-sonnet-4”, “providers”: { “anthropic”: { “apiKey”: “your-api-key” } }, “agents”: { “defaults”: { “sandbox”: { “mode”: “non-main” // Sandbox for groups } } }, “channels”: { “whatsapp”: { “allowFrom”: [“+1234567890”], // Whitelist “groups”: [“*”] // Allow all groups } } }
Should You Adopt Moltbot?
✅ Ideal for:
- Technical developers comfortable with command line and Docker
- Early adopters willing to accept security risks
- Users with dedicated hardware (VPS, Mac Mini)
- People valuing privacy and data control
- Those wanting to automate complex workflows
⚠️ Not yet ready for:
- Non-technical users
- Critical production environments
- Sensitive data without proper isolation
- Those expecting a “plug-and-play” solution
💡 Tip: If you’ve never heard of a VPS (Virtual Private Server), it’s better to wait for the project to mature. Moltbot is still in active development and requires technical expertise for secure use.
The Future of Moltbot
The project is evolving rapidly with upcoming features:
- Native mobile apps for iOS and Android
- Live Canvas: visual workspace driven by the agent
- Voice Wake & Talk Mode: always-on voice interaction
- Multi-agent communication: orchestration of AI agents collaborating on tasks
- Security improvements: enhanced sandboxing, granular permissions
The team is also working on improving the installation experience for less technical users, while maintaining the power and flexibility that made the project successful.
Conclusion
Moltbot (formerly ClawdBot) represents a significant advancement in the field of personal AI assistants. By combining the power of large language models with real action capabilities, the project demonstrates what’s possible when you give AI the “hands” to act in the real world.
However, with this power comes important responsibility. Security risks are real and should not be taken lightly. For now, Moltbot is a tool for technical early adopters who understand the implications of giving an AI such extensive access to their systems.
If you’re a developer and the idea of an AI assistant that can truly manage your digital life intrigues you, Moltbot is worth exploring. Just make sure to set it up on an isolated system, read the security documentation carefully, and proceed with caution.