armour

PreHooks.ai

🛡️ The Firewall for AI Agents - Block dangerous commands before they execute.

curl -fsSL prehooks.ai/install.sh | bash

What it does

PreHooks intercepts tool calls from AI coding assistants and blocks dangerous operations:

• rm -rf / → BLOCKED
• curl ... | bash → BLOCKED
• SSH key exfiltration → BLOCKED
• Crypto miners → BLOCKED

Supported Tools

• Claude Code / Claude Desktop / Claude Cowork
• Cursor
• Windsurf
• Gemini CLI
• GitHub Copilot
• Amp Code
• OpenCode

How it works

1. Pattern matching - Fast regex-based blocking for known threats (~5ms)
2. LLM analysis - Cerebras-powered analysis for ambiguous cases (~100ms)
3. Logging - All tool calls logged for audit

Links

• 🌐 prehooks.ai - Landing page
• 📖 Documentation
• 🐛 Report Issues

---

Armour (MCP Proxy)

Security-enhanced MCP proxy for Claude Code - Block destructive tool calls and audit all operations.

Installation

One-line install (binary bundle)

curl -fsSL https://raw.githubusercontent.com/fuushyn/armour/main/scripts/install-armour.sh | bash

This downloads a prebuilt plugin bundle, installs the marketplace locally, installs the plugin, and enables it by writing to ~/.claude/settings.json. Restart Claude Code to load the plugin.

Overrides:

ARMOUR_RELEASE_TAG=v1.0.3
ARMOUR_RELEASE_URL=https://github.com/fuushyn/armour/releases/download/v1.0.3/armour-plugin-darwin-arm64.tar.gz
ARMOUR_INSTALL_DIR=~/.armour/armour-plugin

Use ARMOUR_MARKETPLACE_SOURCE to fall back to a Git marketplace (for example, https://github.com/fuushyn/armour.git).

Manual install

1. Add marketplace in Claude Code:

   /plugin add-marketplace fuushyn/armour
   

2. Install plugin:

   /plugin install armour
   

3. Restart Claude Code to load the plugin

4. View MCP tools:

   /mcp
   

That's it! All your MCP servers now route through Armour with security policies applied.

Features

• 🔒 Security Policies: Choose from strict, moderate, or permissive policies
• 📊 Multi-Server Support: Aggregate multiple MCP servers with automatic namespacing
• 🎯 Audit Logging: Complete audit trail of all tool calls
• ⚡ Zero Configuration: Automatic server detection and configuration

Configuration

Servers are configured in ~/.armour/servers.json and automatically synced on each session start.

Security Policies

• Strict: Blocks rm*, delete*, drop*, and sampling attempts. Read-only mode.
• Moderate: Blocks destructive operations but allows most normal operations.
• Permissive: No blocking, audit mode only.

License

MIT License - see LICENSE file for details.

---

Made with ❤️ for AI safety