Securing ClawdBot with Cloudflare Tunnel
Learn about the security risks of exposed ClawdBot instances on Shodan and how to secure them using Cloudflare Tunnel.
Securing ClawdBot with Cloudflare Tunnel
ClawdBot is a powerful AI assistant that runs locally on your machine. While it's convenient for connecting to WhatsApp, Telegram, Discord and other messengers, using it with default settings poses security risks. This guide covers the dangers of exposed ClawdBot instances on Shodan and how to secure them using Cloudflare Tunnel.
What is ClawdBot?
ClawdBot is an open-source personal AI assistant based on Claude AI. It runs locally on your machine and connects to various messenger platforms through a Gateway server.
Key Features
- Runs locally (Mac, Windows, Linux, Raspberry Pi)
- Supports WhatsApp, Telegram, Discord, Slack, and more
- WebSocket-based Gateway (default port: 18789)
- Complete data privacy
The Problem: Exposed ClawdBot Instances on Shodan
Searching for port 18789 on Shodan reveals publicly exposed ClawdBot Gateway instances across the internet.
Why is this dangerous?
- Unauthorized Access: Anyone can connect to the Gateway without authentication
- Conversation Leakage: All AI conversations could be exposed
- API Key Theft: Connected Claude API keys could be stolen
- Bot Hijacking: Attackers could take control of your messenger bots
Many users enable external access with --bind 0.0.0.0 for convenience, without setting up firewalls or authentication.
The Solution: Cloudflare Tunnel
Cloudflare Tunnel allows secure access to your server without directly exposing ports to the internet.
Benefits
- No Port Exposure: No need to open port 18789 to the internet
- DDoS Protection: Cloudflare's protection applied automatically
- Access Policies: Control access by email, IP, and more
- Free: Basic features are completely free
Related Posts
Agent Production — From Guardrails to Docker Deployment
Build safe agents with 3-layer Guardrails (Input/Output/Semantic), deploy with FastAPI + Docker. Includes HITL, rate limiting, and production monitoring checklist.
MCP + Multi-Agent — How Agents Share Tools and Collaborate
Solve the N×M integration problem with MCP, build role-based multi-agent teams with CrewAI, and implement agent-to-agent delegation with A2A patterns.
LangGraph in Practice — Reflection Agents and Planning Patterns
Auto-validate agent outputs with Self-Critique and systematically decompose complex tasks with Planning patterns. Implementing Generator-Reflector architecture and Plan-and-Execute.