Back to home

GardWatch

Stop supply chain attacks before they start

The gap between vulnerability disclosure and exploitation is almost instant. GardWatch intercepts malicious and compromised packages at install time, whether from a terminal, CI pipeline, or AI coding assistant.

Free and open sourceWorks with AI agents6 ecosystems supported
claude code — ~/project
Claude Codev2.1.87
Opus 4.6 (1M context) with high effort
Connected · Opus 4.6
GardWatch active

The problem

Your biggest attack surface is what you install

AI coding assistants install packages autonomously. A single hallucinated or typosquatted package name can compromise your entire environment. Traditional scanners only catch known CVEs after the damage is done.

Malware & backdoors

Detects known malicious packages and compromised versions before they reach your codebase.

Typosquatting

Catches packages with names designed to trick developers and AI agents into installing malware.

Dependency confusion

Identifies namespace squatting and suspicious version spikes that signal supply chain attacks.

Homoglyph attacks

Spots packages using non-ASCII characters to disguise themselves as legitimate libraries.

Obfuscated payloads

Deep code scanning with AST analysis detects hidden eval, exec, and network calls in package source.

Unmaintained packages

Flags abandoned or low-trust dependencies based on age, popularity, and OpenSSF Scorecard data.

How it works

Trust scoring, not just CVE lookups

1

Intercept

GardWatch hooks into package installations across your terminal, IDE, and AI assistants. Every dependency is checked before it touches your project.

2

Evaluate

Each package is scored from 0 to 100 using 11 trust checks: malware databases, typosquatting detection, package age, popularity, OpenSSF Scorecard, and more.

3

Block or allow

Safe packages pass through silently. Suspicious packages get flagged with details. Critical threats are blocked instantly, with clear explanations of what was detected.

Where it runs

Everywhere your developers are

AI agents and developers should not have to think about supply chain security. GardWatch runs silently in the background, blocking threats wherever packages get installed.

AI coding assistants

Claude Code, Cursor, and other AI tools can install packages autonomously. GardWatch intercepts every install and blocks compromised dependencies before they land.

IDEs

VS Code and JetBrains extensions surface trust scores inline while you edit lockfiles. See risks before you commit.

CI/CD pipelines

Gate your builds on dependency health. GardWatch exits with a non-zero code on critical findings, blocking compromised packages from shipping.

Terminal

Run scans directly from your terminal. Analyze entire dependency files or check a single package before installing.

Ecosystems

Covers the ecosystems that matter

From Python to .NET, GardWatch scans dependency manifests, lockfiles, and SBOMs across the most targeted package ecosystems.

PyPI
npm
Cargo
Go
Maven
NuGet

Get started

Install GardWatch

Free and open source. Pick the option that fits your workflow.

CLI

Standalone scanner for terminal and CI/CD

$ pip install gardwatch
View setup docs

Claude Code

MCP server for AI agent protection

$ npx @gardera/gardwatch-mcp setup-claude

Requires a Gardera account. Sign up for free

VS Code & Cursor

Gardera Security extension

$ code --install-extension gardera.gardera-vscode

Other MCP clients

Windsurf, Cursor, and more

GardWatch works with any MCP-compatible client. Follow the setup guide for your tool.

Requires a Gardera account. Sign up for free

© 2026 Gardera Security. All rights reserved.