ClawdBot: The Open Source Personal AI Assistant Revolutionizing Local Automation
ClawdBot is an open source personal AI assistant that runs on your own machine. With over 12,000 GitHub stars, it integrates WhatsApp, Telegram, Discord and 50+ services for complete automation.
Announcement
ClawdBot has emerged as one of the most promising open source projects of 2026 in the personal AI assistant space. Developed by Peter Steinberger and a growing community, this TypeScript-based tool allows you to run a complete AI agent on your own machine, connected to your favorite messaging platforms.
With over 12,400 GitHub stars and a rapid release cadence, ClawdBot represents a new generation of AI software: local-first, open source, and genuinely useful for daily tasks.
What is ClawdBot?
ClawdBot is a personal AI assistant that runs locally on your devices (macOS, Linux, Windows via WSL2). Unlike traditional cloud assistants, your data stays on your machine and you maintain complete control.
The tool connects to messaging platforms you already use:
- WhatsApp - Personal messages and groups
- Telegram - Bots and conversations
- Discord - Servers and DMs
- Slack - Professional workspaces
- Signal - Encrypted messaging
- iMessage - Apple ecosystem
- Microsoft Teams - Enterprise collaboration
- Matrix - Decentralized protocol
Local-First Architecture
Gateway and Agent
ClawdBot splits into two main components:
- Gateway: A daemon managing messaging connections and exposing a WebSocket control plane on
localhost:18789 - Agent: The LLM-powered brain (Claude, GPT, local models) that processes requests
DEVELOPERbash# Quick installation npm install -g clawdbot@latest clawdbot onboard --install-daemon
The setup wizard guides users through:
- Gateway configuration
- Messaging channel connections
- Skills (plugins) installation
Persistent Memory
ClawdBot stores its memory as Markdown files, similar to Obsidian:
~/.clawdbot/
├── clawdbot.json # Configuration
├── memory/ # Persistent memory
│ ├── context.md # Current context
│ └── notes/ # Organized notes
└── skills/ # Installed skills
This approach makes it easy to inspect, modify, and version control the assistant's knowledge.
Key Features
Full System Access
ClawdBot can execute shell commands, manipulate files, and control your browser:
DEVELOPERpython# The assistant can: # - Read and write files # - Execute terminal commands # - Control Chrome/Chromium # - Send notifications # - Access geolocation (mobile)
50+ Native Integrations
| Category | Services |
|---|---|
| Productivity | Notion, Obsidian, Gmail, Calendar |
| Development | GitHub, Linear, VS Code |
| Smart Home | Philips Hue, Home Assistant |
| Media | Spotify, Twitter/X |
| AI | Claude, GPT, local models |
Voice Capabilities
On macOS, iOS, and Android, ClawdBot offers:
- Voice Wake: Voice command activation
- Talk Mode: Natural voice conversations
- Transcription: Via Groq Whisper
- Text-to-Speech: Via ElevenLabs
Multi-Session Isolation
For security, non-primary sessions (groups, channels) can be isolated in Docker containers:
DEVELOPERjson{ "sandbox": { "enabled": true, "mode": "docker" } }
Real-World Use Cases
Users report impressive automations:
- Restaurant reservations handled automatically
- Email management with sorting and replies
- Git commits triggered by message
- Calendar alerts with traffic integration
- RSS monitoring to task manager
- Home automation via natural conversation
"ClawdBot is an agentic version of Home Assistant" - User feedback
Recommended Configuration for RAG
For RAG applications, ClawdBot works particularly well with:
DEVELOPERjson{ "model": "claude-opus-4-5-20251101", "provider": "anthropic", "memory": { "type": "markdown", "path": "~/.clawdbot/memory" } }
Documentation recommends Anthropic Pro/Max with Opus 4.5 for long-context capabilities.
Comparison with Alternatives
| Feature | ClawdBot | Cloud Assistants | Zapier |
|---|---|---|---|
| Local execution | Yes | No | No |
| Open source | Yes (MIT) | No | No |
| Native messaging | 10+ | Limited | Via API |
| System access | Full | None | Limited |
| Persistent memory | Markdown | Proprietary | No |
| Cost | LLM tokens | Subscription | Subscription |
Community Reception
Federico Viticci (MacStories) describes ClawdBot as "the ultimate expression of a new generation of malleable software".
One power user reports consuming 180 million Anthropic tokens weekly, illustrating the intensity of possible usage.
"2026 is already the year of personal agents. ClawdBot is the best agentic system I've used since Claude Code itself." - @alex_finn
Installation
Prerequisites
- Node.js 22+
- macOS, Linux, or Windows (WSL2)
- API keys for LLM providers
Quick Install
DEVELOPERbash# Via npm npm install -g clawdbot@latest # Initial setup clawdbot onboard --install-daemon
From Source
DEVELOPERbashgit clone https://github.com/clawdbot/clawdbot cd clawdbot npm install npm run build
Security and Privacy
ClawdBot includes several security mechanisms:
- DM Pairing: Unknown senders receive validation codes
- Allowlists: Fine-grained control over authorized users
- Docker Sandbox: Isolation for non-primary sessions
- macOS TCC: Respect for system permissions
Versions and Releases
The current version (January 2026) is v2026.1.21, with frequent releases adding:
- New provider support (OpenCode Zen, MiniMax)
- CLI/Gateway UX improvements
- Optional plugins with approval gates
- Improved reliability for WhatsApp, Telegram, Signal
Resources
- Official website: clawd.bot
- Documentation: docs.clawd.bot
- GitHub: github.com/clawdbot/clawdbot
- Discord: discord.gg/clawd
- License: MIT
Conclusion
ClawdBot represents a major evolution in personal AI assistants. By combining local execution, rich integrations, and open source architecture, it offers a credible alternative to proprietary cloud assistants.
For teams working on RAG systems, ClawdBot can serve as a deployment interface to expose pipelines through existing messaging platforms, creating a natural and accessible user experience.
The project perfectly illustrates the 2026 trend toward truly useful, customizable, and privacy-respecting AI agents.
Tags
Related Posts
Automatic RAG Evaluation: New Framework Achieves 95% Correlation with Human Judgments
Google Research introduces AutoRAGEval, an automated evaluation framework that reliably assesses RAG quality without human annotation.
TechCorp Solutions: 65% Ticket Reduction with RAG-Powered Support
How TechCorp Solutions reduced support tickets by 65% and improved customer satisfaction by implementing a RAG chatbot for their SaaS platform.
RAG for SMBs: Complete Guide Without a Data Team
Deploy a performant RAG system in your SMB without advanced technical skills: no-code solutions, controlled budget, and fast ROI.