As AI agents become more capable, they become prime targets for attackers. OpenClaw provides the brains, but exposing an unprotected agent to the wild is a massive risk. Prompt injections, secret exfiltration, and unexpected behavior can compromise your systems in seconds.
KnightClaw is the solution.
We built the ultimate zero-trust security extension for OpenClaw. Built on a philosophy of "Never Trust, Always Verify," KnightClaw intercepts every request and response, passing it through a rigorous, high-speed 4-layer defense system. If your agent is the brain, KnightClaw is the immune system.
KnightClaw operates as a synchronous interceptor within OpenClaw. Before any tool runs or any model generates text, KnightClaw scrutinizes the data. It uses blazing-fast regex patterns, boundary detection, and a local semantic embedding model—happening entirely on-device, completely private.
flowchart TD
User([Attacker / User]) -->|Payload| KC{KnightClaw Engine}
subgraph KC_Layers [The 4 Layers of Defense]
L1[🛡️ Guard 2.0<br>Input Validation]
L2[📋 Audit Logs<br>Tamper-Proof]
L3[🔐 Vault<br>Data Encryption]
L4[🚨 Lockdown<br>Kill Switch]
L1 --> L2
L2 --> L3
L3 --> L4
end
KC -->|Intercepts| KC_Layers
KC_Layers -->|Blocked!| Lockdown[System Locked]
KC_Layers -->|Safe!| O[OpenClaw Agent Core]
O -->|Executes| Task[Run Task / Tool]
| Layer | Feature | Why It Matters | Status |
|---|---|---|---|
| 1 | 🛡️ Guard 2.0 | Blocks prompt injections, malicious inputs, homoglyphs, and jailbreaks using an advanced hybrid heuristics and semantic engine. | Active |
| 2 | 📋 Logs | A secure, hash-chained, immutable audit trail. Every action is recorded—tamper-proof. | Active |
| 3 | 🔐 Vault | Encrypts all your secrets at rest (AES-256-GCM), ensuring keys are never exposed in plaintext. | Active |
| 4 | 🚨 Lockdown | The ultimate emergency kill switch. Detects high-velocity attacks and instantly blocks all operations. | Armed |
Our guarantee: If one layer fails, 3 others are still guarding the gate.
Getting started is designed to be frictionless. We currently support local installation via the knightclaw folder.
# 1. Navigate to the plugin directory
cd /path/to/knightclaw
# 2. Install dependencies & build
npm install
npm run build
# 3. Load the plugin into OpenClaw (Adds us to your config)
npm run loadThat's it. KnightClaw is now actively protecting your OpenClaw instance with maximum security.
KnightClaw comes with a powerful CLI packed right inside OpenClaw. Control your security posture with ease.
Check the real-time status of all your security layers.
openclaw knight statusDisable features only when absolutely necessary (requires confirmation).
openclaw knight enable guard
openclaw knight disable guard --confirmView your tamper-proof, redacted security history.
openclaw knight logs # View recent entries
openclaw knight logs -f # Follow logs in real-time
openclaw knight logs verify # Cryptographically verify integrity Under attack? Pull the plug. Instantly prevents OpenClaw from taking any action.
openclaw knight lockdown on # INSTANT kill switch
openclaw knight lockdown off -y # Resume operations- Default = Most Secure — We lock everything down out of the box. You slowly loosen it (if needed).
- Never Trust, Always Verify — Even localhost traffic gets deeply inspected.
- Explicit Consent required — Want to turn off a security feature? You have to
--confirmit, and we log that decision. - Simple Over Complex — Security features you don't understand are security vulnerabilities. We keep it incredibly simple.
Built for Enterprise. Designed for Everyone. 🛡️