DepShield MCP
Acts as a security checkpoint for AI coding agents by intercepting package installations to verify existence, check against CVE databases, and block vulnerable or hallucinated dependencies before they reach your codebase. Provides seven security tools including pre-install gates, full project audits, safe version recommendations, and deep transitive dependency scanning for npm and PyPI packages.
README
<p align="center"> <img src="https://img.shields.io/badge/๐ก๏ธ_DepShield-MCP_Server-blue?style=for-the-badge" alt="DepShield MCP" /> </p>
<h3 align="center">Real-time dependency security for AI coding agents</h3>
<p align="center"> <a href="https://www.npmjs.com/package/depshield-mcp"><img src="https://img.shields.io/npm/v/depshield-mcp?style=flat-square&color=cb3837&label=npm" alt="npm version" /></a> <a href="https://www.npmjs.com/package/depshield-mcp"><img src="https://img.shields.io/npm/dm/depshield-mcp?style=flat-square&color=blue&label=downloads" alt="npm downloads" /></a> <a href="https://github.com/devanshkaria88/depshield-mcp/blob/main/LICENSE"><img src="https://img.shields.io/github/license/devanshkaria88/depshield-mcp?style=flat-square" alt="license" /></a> <a href="https://github.com/devanshkaria88/depshield-mcp"><img src="https://img.shields.io/github/stars/devanshkaria88/depshield-mcp?style=flat-square" alt="GitHub stars" /></a> <img src="https://img.shields.io/badge/node-%3E%3D22-brightgreen?style=flat-square" alt="node version" /> </p>
<p align="center"> DepShield intercepts every package your agent tries to install, verifies it exists, checks it against CVE databases, and blocks vulnerable or hallucinated dependencies โ <strong>before they touch your codebase</strong>. </p>
<p align="center"> Works with <strong>Cursor</strong> ยท <strong>Claude Code</strong> ยท <strong>Windsurf</strong> ยท <strong>Antigravity</strong> ยท <strong>Cline</strong> ยท any MCP-compatible IDE </p>
๐ค Why
AI coding agents pull packages from stale training data. They install outdated versions with known CVEs. They hallucinate package names that don't exist, opening the door to typosquatting attacks. They default to deprecated libraries when better alternatives exist.
There's no checkpoint between "agent decides to use a package" and "package lands in your project." DepShield is that checkpoint.
โก What It Does
Seven security tools exposed over the Model Context Protocol:
| Tool | What it does |
|---|---|
check_dependency |
Pre-install gate โ verifies a package exists on the registry and has no known CVEs. Your agent calls this before every install. |
audit_project |
Scans an entire package.json or requirements.txt and returns a full vulnerability audit report. |
find_safe_version |
Finds the newest version of a package with zero known vulnerabilities. |
get_advisory_detail |
Deep dive into a specific CVE/GHSA โ full description, affected versions, fix info, references. |
check_npm_health |
Package health report card: weekly downloads, last publish date, maintainer count, license, deprecation status. Scored 0โ100. |
suggest_alternative |
Finds better packages when one is vulnerable, deprecated, or abandoned. |
deep_scan |
Scans a package's transitive dependency tree for vulnerabilities, typosquats, and suspicious patterns. |
Plus a depshield://status resource and a security_review prompt template for guided full-project audits.
Zero API keys required. DepShield uses free, open APIs: npm Registry, OSV.dev (Google's open vulnerability database), and PyPI.
๐ Quick Start
Option 1: npx (no install)
Add to your IDE's MCP configuration and you're done. No global install needed.
Option 2: Clone and build
git clone https://github.com/devanshkaria88/depshield-mcp.git
cd depshield-mcp
npm install
npm run build
๐ IDE Setup
<details> <summary><strong>Cursor</strong></summary>
Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json (project-level):
{
"mcpServers": {
"depshield": {
"command": "npx",
"args": ["-y", "depshield-mcp"]
}
}
}
Or if you cloned the repo locally:
{
"mcpServers": {
"depshield": {
"command": "node",
"args": ["/absolute/path/to/depshield-mcp/dist/index.js"]
}
}
}
Optionally, copy .cursor/rules/dep-shield.mdc into your project's .cursor/rules/ directory. This rule forces the Cursor agent to call check_dependency before every package install.
</details>
<details> <summary><strong>Claude Code</strong></summary>
claude mcp add depshield -- npx depshield-mcp
Or from a cloned repo:
claude mcp add depshield -- node /absolute/path/to/depshield-mcp/dist/index.js
</details>
<details> <summary><strong>Windsurf</strong></summary>
Add to your Windsurf MCP configuration:
{
"mcpServers": {
"depshield": {
"command": "npx",
"args": ["-y", "depshield-mcp"]
}
}
}
</details>
<details> <summary><strong>Any MCP-compatible tool</strong></summary>
DepShield uses stdio transport. Any tool that supports MCP over stdio can use it:
npx depshield-mcp
</details>
๐ก Usage Examples
Once connected, your AI agent has access to all seven tools. Try these prompts:
<table> <tr> <td width="50%">
Pre-install check (automatic with .mdc rule)
"Add lodash for deep cloning"
Agent calls check_dependency โ finds CVE โ auto-upgrades to safe version
</td> <td width="50%">
Block hallucinated packages
"Install react-super-utils-pro for state management"
Agent calls check_dependency โ package doesn't exist โ blocks install, suggests alternatives
</td> </tr> <tr> <td>
Full project audit
"Run a security audit on this project's dependencies"
Agent calls audit_project on package.json โ returns full vulnerability report
</td> <td>
Package health check
"Is this package well maintained?"
Agent calls check_npm_health โ returns health score (0โ100) with breakdown
</td> </tr> <tr> <td>
Supply chain deep scan
"Deep scan express for transitive vulnerabilities"
Agent calls deep_scan โ scans dependency tree, flags suspicious patterns
</td> <td>
Advisory deep dive
"Tell me more about GHSA-jf85-cpcp-j695"
Agent calls get_advisory_detail โ returns full CVE details, remediation info
</td> </tr> </table>
๐ Tools Reference
<details> <summary><code>check_dependency</code> โ Pre-install gate</summary>
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
name |
string | yes | โ | Package name (e.g., lodash, express) |
version |
string | no | latest | Specific version to check |
ecosystem |
npm | pypi |
no | npm |
Package ecosystem |
Returns: โ
SAFE, โ ๏ธ VULNERABLE (with fix version), ๐ซ BLOCKED (doesn't exist), or โ ๏ธ CANNOT VERIFY (registry unreachable).
</details>
<details> <summary><code>audit_project</code> โ Full manifest audit</summary>
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
filePath |
string | yes | โ | Path to package.json or requirements.txt |
includeDevDependencies |
boolean | no | true |
Include devDependencies in scan |
Returns: Full audit report with summary stats, per-dependency vulnerability breakdown, severity counts, and risk verdict.
</details>
<details> <summary><code>find_safe_version</code> โ Safe version finder</summary>
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
name |
string | yes | โ | Package name |
ecosystem |
npm | pypi |
no | npm |
Package ecosystem |
Returns: The newest stable version with zero known vulnerabilities, with comparison of all checked versions.
</details>
<details> <summary><code>get_advisory_detail</code> โ CVE/GHSA deep dive</summary>
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
vulnId |
string | yes | โ | Vulnerability ID (e.g., GHSA-jf85-cpcp-j695, CVE-2021-23337) |
Returns: Full advisory with summary, severity, CVSS score, affected versions, fix versions, and reference links.
</details>
<details> <summary><code>check_npm_health</code> โ Package health scoring</summary>
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
name |
string | yes | โ | npm package name |
Returns: Health report card (0โ100 score) based on: publish recency, weekly downloads, license, repository presence, deprecation status, and maintainer count.
</details>
<details> <summary><code>suggest_alternative</code> โ Alternative package finder</summary>
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
name |
string | yes | โ | Package to find alternatives for |
reason |
string | no | โ | Why an alternative is needed |
Returns: Top 3 alternative packages with downloads, npm score, and publish date.
</details>
<details> <summary><code>deep_scan</code> โ Transitive dependency scanner</summary>
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
name |
string | yes | โ | Package to deep scan |
version |
string | no | latest | Specific version to scan |
depth |
1 | 2 |
no | 1 |
1 = direct deps, 2 = deps of deps |
Returns: Dependency tree with vulnerability flags, suspicious pattern detection (newly added deps, nonexistent packages, low-download typosquat candidates), and risk verdict.
</details>
๐ Cursor Rule (Optional)
The .cursor/rules/dep-shield.mdc file included in this repo forces the Cursor agent to automatically call check_dependency before every package install. Copy it to your project:
mkdir -p .cursor/rules
cp node_modules/depshield-mcp/.cursor/rules/dep-shield.mdc .cursor/rules/
This turns DepShield from a tool the agent can use into a gate the agent must pass through.
๐งช Testing
MCP Inspector (interactive UI for testing tools):
npm run inspect
Or:
npx @modelcontextprotocol/inspector node dist/index.js
Raw stdio (quick verification):
printf '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0.0"}}}\n{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"check_dependency","arguments":{"name":"lodash","version":"4.17.20"}}}\n' | node dist/index.js 2>/dev/null
๐ APIs Used
All free, no API keys required:
| API | What it provides |
|---|---|
| npm Registry | Package existence, versions, metadata, search |
| npm Downloads API | Weekly download counts |
| OSV.dev | Open source vulnerability database (Google) |
| PyPI | Python package metadata |
๐ Project Structure
depshield-mcp/
โโโ src/
โ โโโ index.ts # MCP server entry โ tool/resource/prompt registration
โ โโโ cache.ts # In-memory TTL cache (5 min)
โ โโโ utils.ts # Severity parsing, version sorting, fetch helpers
โ โโโ apis/
โ โ โโโ npm-registry.ts # npm registry, search, downloads
โ โ โโโ osv.ts # OSV.dev vulnerability queries
โ โ โโโ pypi-registry.ts # PyPI JSON API
โ โโโ tools/
โ โโโ check-dependency.ts # Pre-install gate
โ โโโ audit-project.ts # Full manifest audit
โ โโโ find-safe-version.ts # Safe version finder
โ โโโ get-advisory-detail.ts # CVE/GHSA deep dive
โ โโโ check-npm-health.ts # Package health scoring
โ โโโ suggest-alternative.ts # Alternative package finder
โ โโโ deep-scan.ts # Transitive dependency scanner
โโโ .cursor/rules/
โ โโโ dep-shield.mdc # Cursor agent rule (alwaysApply)
โโโ package.json
โโโ tsconfig.json
โโโ LICENSE
โโโ README.md
๐ Requirements
- Node.js 22 or later
- An MCP-compatible AI coding tool
๐ค Contributing
Contributions are welcome. Please open an issue first to discuss what you'd like to change.
Areas where help would be appreciated:
- Additional ecosystem support (Cargo, Go modules, Maven)
- Improved CVSS vector string parsing
- Integration tests with mocked API responses
- GitHub Actions CI pipeline
๐ License
MIT โ see LICENSE for details.
<p align="center"> Built by <a href="https://github.com/devanshkaria88">Devansh Karia</a> </p>
<p align="center"> <a href="https://buymeacoffee.com/devanshkaria88"> <img src="https://cdn.buymeacoffee.com/buttons/v2/default-yellow.png" alt="Buy Me A Coffee" height="50" width="217" /> </a> </p>
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.