135,000+ OpenClaw instances are running without audit logs, access controls, or PII protection. LogionOS Shield adds enterprise-grade compliance in one command. Self-hardened. Zero dependencies. Open source. Free forever.
Shield's protection is measured against industry-standard security benchmarks. No marketing claims — just numbers.
Shield intercepts at four critical lifecycle points in the OpenClaw gateway — before and after every AI interaction.
Scans user messages for PII, credentials, and prompt injection before they reach the AI
Detects ClawJacked attack signatures and malicious context injection in system prompts
Controls which tools the agent can use, scans arguments for dangerous operations
Filters AI responses for data leaks and automatically adds regulatory disclaimers
Comprehensive coverage of OpenClaw's known vulnerabilities and compliance gaps
Tamper-evident SHA-256 hash chain audit trail. Every compliance decision is recorded with full context, exportable as JSON/CSV. Compatible with SOC 2 and ISO 27001 evidence requirements.
Dedicated Prompt Guard with 8+ ClawJacked attack signatures, control character detection, and context file injection scanning.
12+ PII patterns detected in real-time: SSN, credit cards, API keys, JWT tokens, AWS credentials, private keys, passwords, My Number (JP), and more.
Tool Guard enforces allowlist/denylist policies per tool, scans arguments for destructive operations (rm -rf, DROP TABLE, SQL injection), and blocks unauthorized actions.
Outbound Guard scans every AI response for PII before delivery. Detects leaked credentials, personal data, and adds regulatory disclaimers for medical/legal/financial topics.
Kill Switch instantly blocks all AI interactions across all channels. Switchable enforcement modes: Monitor → Enforce → Strict, with per-guard toggle controls.
Tool Guard pre/post scanning detects dangerous operations from malicious skills at runtime. Combined with denylist policies, prevents most skill-based attacks.
Blocks shell execution, filesystem deletion, and SQL injection patterns in tool arguments. Detects cloud metadata access and pipe-based download attacks.
Real security breaches and crypto theft cases involving OpenClaw and AI agents. Each one preventable by Shield.
Shield ships with an embedded web dashboard — no separate deployment needed. Real-time monitoring at localhost:18789/logionos/
A security tool that can be compromised is worse than no security at all. Shield is hardened against 6 categories of self-targeted attacks.
In enforce/strict mode, Shield refuses to connect to non-HTTPS API endpoints. Prevents MITM attacks on the compliance decision channel.
requireTls() redirect: errorDashboard uses strict Content Security Policy, per-session CSRF tokens, X-Frame-Options: DENY, and constant-time token comparison.
CSP X-CSRF-Token DENYIf Shield crashes or becomes unresponsive, a watchdog automatically blocks all requests. No silent fail-open — your data stays protected.
heartbeat 10s timeout 30sNFKC Unicode normalization + zero-width character stripping before every scan. Prevents homoglyph and invisible character bypass attacks.
NFKC \u200B stripInput truncated at 50K characters. Regex iterations capped at 500 per pattern. Crafted payloads cannot cause catastrophic backtracking.
MAX_SCAN 50K iter < 500API endpoint validated at startup. Cloud metadata endpoints blocked. Path locked to /v1/*. Redirects rejected. No open-redirect exploitation.
169.254.x block /v1/* onlyDashboard API endpoints rate-limited to 120 req/min per IP. Prevents brute-force attacks and denial-of-service on the management interface.
120/min/IPShield computes a SHA-256 hash of its own critical modules at startup. Any tampering is detected and logged to the audit trail.
SHA-256 computeModuleHash()Shield uses only Node.js built-in modules (crypto, fs, path). Zero npm dependencies means zero supply chain attack surface.
0 deps node:crypto onlyOne command. No configuration files to edit. No separate services to deploy.
# Install the plugin
$ openclaw plugins install @logionos/openclaw-shield
# Restart gateway
$ openclaw restart
# Open dashboard
$ open http://localhost:18789/logionos/Every message without Shield is a compliance risk. Install now, monitor first, enforce when ready.