Vulnerability Research

1,673 Moltbot (Formerly Clawdbot) AI Gateways Exposed: Inside Our Security Research

On a routine Shodan sweep of Moltbot (formerly Clawdbot) AI gateways, we expected to find a few misconfigured reverse proxies. Instead, we uncovered 1,673 exposed Moltbot servers, 55 JSON-RPC methods, and more than 1,000 instances with no authentication at all. This report explains how a single line of JavaScript turned Moltbot's "localhost trust" model into a widespread security weakness across the AI gateway ecosystem.

This report presents an in-depth security analysis of Moltbot (formerly Clawdbot), focusing on exposed gateways, unauthenticated access, and critical vulnerabilities affecting real-world deployments. We identified 1,673 publicly exposed Moltbot servers, documented 55 JSON-RPC methods (including new high-impact actions), and validated three critical CVEs through live exploitation scenarios.

Sable Security Research Team
January 28, 2026
0 views
12 min read

TL;DR: Moltbot/Clawdbot Security Findings Summary

1,673
Servers Found
via Shodan
1,000+
Without Auth
publicly accessible
55
RPC Methods
attack surface
3 CVEs
Critical Vulns
CVSS 8.8-9.6

Affected data: OpenAI/Anthropic API keys, SSH private keys, WhatsApp/Telegram/Slack conversations, Discord webhooks, AWS credentials, and complete user chat histories with AI assistants.

Methodology: Scanning Moltbot Gateways and Exposed Admin Ports

Our investigation began with a simple Shodan query targeting Moltbot and Clawdbot gateway signatures. The results revealed a widespread exposure problem affecting servers across 13+ countries.

shodan-recon.sh
$ shodan search "moltbot OR clawdbot" --fields ip_str,port,org
Searching Shodan...
Total Results: 1,673
Top ports: 18789 (986), 5353 (1187), 8080 (hundreds)
Top countries: US (334), DE (249), UAE (174), FI (146)
Warning: 1,000+ servers appear to lack authentication

The high concentration in certain hosting providers (Hetzner: 593, DigitalOcean: 217) suggests many deployments are developer/hobbyist setups following tutorials without security hardening.

The Localhost Bypass Vulnerability in Moltbot/Clawdbot

The root cause is elegant in its simplicity and devastating in its impact. We analyzed the npm package source code and found the exact vulnerable logic:

message-handler.js:312
// The vulnerability that exposed everything
const isLocalClient = !hasUntrustedProxyHeaders && isLocalGatewayAddress(clientIp);
const nonceRequired = !isLocalClient;// <-- localhost = no auth required
if (nonceRequired && !providedNonce) {
close(1008, "device nonce required");
}
The Assumption vs Reality

Assumption: "If you can reach localhost, you're already on the machine."
Reality: Docker networks, reverse proxies (nginx, Cloudflare, Caddy), and cloud deployments routinely make "localhost" reachable from the internet.

// Attack flow when reverse proxy is misconfigured
┌──────────────┐     ┌─────────────────┐     ┌─────────────────┐
│   Attacker   │ ──► │ nginx/Cloudflare│ ──► │ Moltbot Gateway │
│  (Internet)  │     │ (Reverse Proxy) │     │  (Port 18789)   │
└──────────────┘     └─────────────────┘     └─────────────────┘
                              │
                    clientIp = 127.0.0.1
                    isLocalClient = true
                    nonceRequired = false
                              │
                    ┌─────────▼─────────┐
                    │   FULL ACCESS      │
                    │   No auth needed  │
                    └───────────────────┘

Findings: 55 JSON-RPC Methods Exposed in Moltbot Gateways

We extracted and enumerated all exposed JSON-RPC methods from the gateway's JavaScript bundle. The attack surface is extensive, with methods enabling everything from credential theft to remote code execution.

CRITICAL - Immediate Compromise
system.runArbitrary shell command execution
config.getFull configuration with API keys
skills.installInstall arbitrary code (RCE vector)
cron.addSchedule persistent backdoors
exec.approval.approveBypass execution approvals
HIGH - Data Exfiltration
chat.historyRead all private conversations
chat.sendSend messages as the user
sessions.listEnumerate active sessions
logs.tailRead system and application logs
agents.listList configured AI agents

Additional methods discovered include session management, channel operations, node administration, and operator permissions - totaling 55 distinct attack vectors.

Live Testing: Verifying Moltbot Server Exposures

We tested 4 servers that appeared active on Shodan, connecting via WebSocket to verify the exposure. All accepted connections and responded with challenge-response authentication prompts.

ServerPortLocationFinding
107.174.96.24218789🇺🇸USWebSocket active, challenge-response enabled
139.198.181.2349002🇨🇳CNWebSocket active, accepts connections
175.24.206.2528081🇨🇳CNWebSocket active, RPC enumerable
175.24.206.25218789🇨🇳CNWebSocket active, multi-port exposure
Responsible Testing

Our testing was limited to connection verification and method enumeration. No user data was accessed, exfiltrated, or stored.

Attack Techniques Tested Against Moltbot Servers

We systematically tested multiple attack vectors to understand what works and what doesn't. This helps defenders prioritize mitigations.

Nonce Prediction
Attempted to predict UUIDv4 challenge nonces
MITIGATED
Origin Header Bypass
Spoofed Origin headers to bypass CORS
MITIGATED
X-Forwarded-For Injection
Injected localhost via proxy headers
MITIGATED
Subprotocol Confusion
Tested WebSocket subprotocol handling
PARTIAL
Default Token Bruteforce
Tested common default credentials
MITIGATED
Parameter Pollution
Attempted JSON parameter injection
MITIGATED
Race Condition
Concurrent auth requests for state confusion
MITIGATED
Timing Attack
Measured response times for info leakage
MITIGATED
Method Case Confusion
Tested case sensitivity of RPC methods
MITIGATED
Reverse Proxy Bypass
Exploited misconfigured nginx/Cloudflare
VULNERABLE

Impact: Data Leaks from AI Agent Gateways

When a Moltbot/Clawdbot server is compromised, attackers gain access to sensitive data stored in configuration files, chat histories, and connected integrations.

API Keys
  • OpenAI API keys (sk-...)
  • Anthropic API keys (sk-ant-...)
  • AWS credentials
  • SSH private keys
Private Conversations
  • WhatsApp chat history
  • Telegram messages
  • Slack channels & DMs
  • Discord server messages
  • Signal pairing credentials
Bot Tokens
  • Telegram bot tokens
  • Slack OAuth credentials
  • Discord webhooks
  • Custom integrations
System Access
  • Shell command execution
  • Cron job manipulation
  • File system access
  • Root privileges (some cases)

CVE-2025-49596, CVE-2025-6514 and CVE-2025-52882 in Moltbot Ecosystems

CVE-2025-6514
Command Injection to RCE
CVSS 9.6

The system.run RPC method allows arbitrary command execution. Combined with the localhost bypass, an attacker can achieve root-level access in a single unauthenticated request.

Vector: Network / No Auth / No Interaction
CVE-2025-49596
Unauthenticated System Compromise
CVSS 9.4

WebSocket API exposed without network-level restrictions. Default configurations allow remote connections, and the localhost trust assumption is bypassed via reverse proxy.

Vector: Network / No Auth / No Interaction
CVE-2025-52882
Arbitrary File Access + Code Execution
CVSS 8.8

Configuration files readable via config.get expose API keys and tokens. Combined with skills.install, attackers can achieve persistent code execution.

Vector: Network / Low Privilege / No Interaction

Findings: 1,673 Exposed Moltbot Servers - Geographic Distribution

By Country

🇺🇸United States
334
🇩🇪Germany
249
🇦🇪UAE
174
🇫🇮Finland
146
🇸🇬Singapore
54
🇨🇳China
50
🌍Others
666

By Hosting Provider

Hetzner Online GmbH
593
DigitalOcean
217
OVH SAS
31
HostPapa
10
Linode
10

High concentration in Hetzner suggests developer/hobbyist deployments without security review.

Mitigation: How to Secure Moltbot and Similar AI Gateways

If you run Moltbot/Clawdbot, take these steps immediately to secure your deployment:

1Check Your Exposure
shodan search "clawdbot-gw" --fields ip_str,port | grep YOUR_IP
2Enable Authentication
export REQUIRE_AUTH=true
export AUTH_TOKEN=$(openssl rand -hex 32)
3Bind to Localhost Only
# In config
gateway:
  host: 127.0.0.1  # NOT 0.0.0.0
  port: 18789
4Configure Trusted Proxies
# If using reverse proxy
gateway:
  trustedProxies:
    - 10.0.0.0/8
    - 172.16.0.0/12
5Rotate All Credentials
# Immediately rotate:
- All LLM API keys
- Bot tokens
- OAuth credentials
- SSH keys if exposed

Disclosure Timeline

Late 2025
Clawdbot gains viral adoption (60,000+ GitHub stars)
Jan 15, 2026
Initial reports of exposed servers surface
Jan 20, 2026
CVEs officially assigned and published
Jan 27, 2026
Anthropic requests project rename to Moltbot
Jan 28, 2026
Sable Security completes comprehensive research
Jan 28, 2026
This report published

Is Your AI Gateway Secure?

We found these vulnerabilities through systematic research. The same methodology can audit your specific deployment before attackers find it.

This research was conducted following responsible disclosure practices. All testing was limited to benign enumeration. No user data was accessed or stored.