HomeDocsPricingDownload

Quick Start

Get ClaudeArmor running in under 60 seconds. Install, scan, done.

# Install via npm $ npx claudearmor@latest # Scan the current directory $ claudearmor scan . # Scan a specific repo before cloning $ claudearmor reposcan https://github.com/user/project
💡 Tip: Add --json to any scan command to output results as JSON for CI/CD integration.

Installation

ClaudeArmor runs on macOS, Windows, and Linux. Choose your preferred method:

MethodCommand
npm / npxnpx claudearmor@latest
Homebrewbrew install claudearmor
pippip install claudearmor
Dockerdocker run ghcr.io/claudearmor/scanner scan .
DesktopDownload page →

Your First Scan

Point ClaudeArmor at any directory, file, or URL and it will run all enabled engines automatically.

$ claudearmor scan ./my-project 🔍 Scanning 1,247 files across 8 engines... ┌────────────────┬────────┬──────────┐ │ Engine │ Status │ Findings │ ├────────────────┼────────┼──────────┤ │ KeyLeak │ ✅ │ 2 high │ │ HeuristicGuard │ ✅ │ 0 │ │ MalwareShield │ ✅ │ 0 │ │ McpScan │ ✅ │ 1 medium │ │ MicroSandbox │ ✅ │ 0 │ │ RepoScan │ ✅ │ 0 │ │ ToolScan │ ✅ │ 0 │ │ ThreatReport │ ✅ │ AI ready │ └────────────────┴────────┴──────────┘ ⚠️ 3 findings detected. Run 'claudearmor report' for details.

KeyLeak Engine

Detects 40+ API key & credential classes using high-precision entropy & structural heuristics. Supports AWS, GCP, Stripe, OpenAI, GitHub, Supabase, Vercel, and more.

Customising Signatures

Add custom enterprise token rules to ~/.claudearmor/keyleak.yml:

signatures: - name: "My Internal Token" prefix: "MYAPP_" severity: high description: "Internal application token"

HeuristicGuard Engine

Behavioral and signature-based payload detection powered by ClaudeArmor's proprietary neural threat feed, updated daily against emerging AI agent exploits.

Use the --custom-rules flag to load organization-specific policies:

$ claudearmor scan . --custom-rules ./org-policies/

MalwareShield Engine

Multi-layer binary and script threat intelligence engine for trojan, dropper, and supply-chain malware detection. Signatures synchronize automatically via encrypted cloud feed.

McpScan Engine

Validates Model Context Protocol tool manifests for suspicious capabilities: network access, file writes, credential requests, and obfuscated tool descriptions.

MicroSandbox Engine

Executes untrusted code in a zero-trust isolated micro-sandbox with strict capability boundaries. Monitors for unauthorized filesystem access, network exfiltration, and child process spawning.

RepoScan Engine

Analyzes remote repositories before cloning. Checks commit history, contributor trust scores, dangerous CI/CD configs, and known malicious patterns.

ToolScan Engine

Static & behavioral analysis of MCP tool source code. Detects obfuscated payloads, dynamic execution chains, and hidden command injection vectors.

AI Threat Reports

Uses ClaudeArmor's proprietary Forensic AI Engine to generate executive explanations of detected threats, including risk assessment, attack scenarios, and step-by-step remediation solutions.

⚡ Included in All Defense Passes: Activate a 3-Month, 6-Month, or 1-Year Pass to unlock full Forensic AI Reports. See Defense Passes →

Configuration

Create .claudearmor.yml in your project root:

engines: keyleak: true heuristics: true malwareshield: true mcpscan: true sandbox: false # disabled for CI speed reposcan: true toolscan: true exclude: - node_modules - .git - "*.lock" severity_threshold: medium output: json

CLI Reference

CommandDescription
scan <path>Scan files or directories
reposcan <url>Scan remote repo before cloning
reportShow last scan report
quarantine listList quarantined files
quarantine restore <id>Restore a quarantined file
updateUpdate signatures and engines
configShow current configuration

API

ClaudeArmor exposes a local HTTP API on port 41927 for integration with editors and CI pipelines.

$ curl http://localhost:41927/api/scan \ -X POST \ -H "Content-Type: application/json" \ -d '{"path": "./my-project", "engines": ["keyleak", "heuristics"]}'