audit-mcp-cli
audit-mcp-cli
Documentation
audit-mcp-cli
English | 中文
A lightweight dependency vulnerability audit tool for Node.js projects. Supports CLI and MCP Server modes, covers npm and pnpm projects, and generates structured Markdown/HTML reports with full dependency chains.
Features
- Full dependency chains — traces the complete path from your package.json to each vulnerable package
- npm + pnpm support — auto-detects package manager by lockfile
- Remote GitHub audit — audit any public or private repo without cloning
- MCP Server — integrates with AI coding assistants (Claude, Cursor, etc.)
- Markdown / HTML reports — clean, structured reports sorted by severity
- CI gate — `--fail-on` exit code for CI/CD pipelines
- Ignore mechanism — suppress accepted vulnerabilities with expiration dates
- Severity filtering — show only vulnerabilities above a threshold
Install
# Run directly
npx audit-mcp-cli
# Or install globally
npm install -g audit-mcp-cliRequires Node.js >= 18.
Usage
# Audit current directory
audit-mcp-cli
# Specific project path
audit-mcp-cli --path /path/to/project
# Remote GitHub repo (branch)
audit-mcp-cli --remote github:facebook/react --ref main
# Remote GitHub repo (tag)
audit-mcp-cli --remote github:facebook/react --ref v18.2.0
# Remote GitHub repo (commit SHA)
audit-mcp-cli --remote github:facebook/react --ref abc123def
# HTML report
audit-mcp-cli --format html --output report.html
# CI: fail if high+ severity vulnerabilities found
audit-mcp-cli --fail-on high
# Severity filtering (only show high and critical)
audit-mcp-cli --severity highCLI Options
| Option | Description | Default |
|---|---|---|
| `--path ` | Local project path | `process.cwd()` |
| `--remote ` | Remote repo: `github:owner/repo` or `https://github.com/owner/repo` | — |
| `--ref ` | Git ref (branch name / tag / commit SHA) | `main` |
| `--token ` | GitHub personal access token (for private repos) | `GITHUB_TOKEN` env |
| `--format ` | Report format: `md` or `html` | `md` |
| `--output ` | Output file path | `audit-report.md` or `.html` |
| `--severity ` | Minimum severity to display: `low` / `moderate` / `high` / `critical` | `low` |
| `--fail-on ` | CI fail threshold — exit 1 if vulnerabilities at this level or above exist | — |
| `--mcp` | Start as MCP Server | — |
| `--lang ` | Language: `en` or `zh-CN` | Auto-detect from system |
`--fail-on` exit codes
| Value | Exits 1 when |
|---|---|
| `critical` | Any critical vulnerability found |
| `high` | Any high or critical found |
| `moderate` | Any moderate, high, or critical found |
| `low` | Any vulnerability found |
| *(not set)* | Always exits 0 |
MCP Server
Run as an MCP stdio server for AI assistants:
audit-mcp-cli --mcpClaude Desktop
Basic (local projects & public repos):
{
"mcpServers": {
"audit-mcp-cli": {
"command": "npx",
"args": ["-y", "audit-mcp-cli", "--mcp"]
}
}
}With GitHub token (private repos / avoid rate limits):
{
"mcpServers": {
"audit-mcp-cli": {
"command": "npx",
"args": ["-y", "audit-mcp-cli", "--mcp"],
"env": {
"GITHUB_TOKEN": "ghp_xxxx"
}
}
}
}Cursor
Add to `.cursor/mcp.json`:
Basic (local projects & public repos):
{
"mcpServers": {
"audit-mcp-cli": {
"command": "npx",
"args": ["-y", "audit-mcp-cli", "--mcp"]
}
}
}With GitHub token (private repos / avoid rate limits):
{
"mcpServers": {
"audit-mcp-cli": {
"command": "npx",
"args": ["-y", "audit-mcp-cli", "--mcp"],
"env": {
"GITHUB_TOKEN": "ghp_xxxx"
}
}
}
}Tool: `audit_dependencies`
The MCP server exposes one tool that supports both local and remote auditing:
| Parameter | Description |
|---|---|
| `projectPath` | Local project path |
| `remoteRepo` | Remote repo: `github:owner/repo` |
| `ref` | Git ref (branch / tag / SHA) |
| `token` | GitHub token (for private repos, or use `GITHUB_TOKEN` env) |
| `format` | `md` or `html` |
| `severity` | Minimum severity filter |
| `outputPath` | Custom output file path |
Returns: report file path + structured vulnerability details (CVSS, dependency chains, fix suggestions).
> Token is optional. Local project auditing never requires a token. Remote public repos work without a token (60 requests/hour). Only private repos require a GitHub token.
Ignore Mechanism
Create `.audit-mcp-cli-ignore.json` in your project root to suppress accepted vulnerabilities:
{
"ignore": [
{
"packageName": "minimist",
"advisorySource": 1179,
"reason": "Accepted risk, limited impact in our usage",
"expiresAt": "2025-12-31T00:00:00Z"
}
]
}- `packageName` — match all advisories for this package, or combine with `advisorySource` for exact match
- `expiresAt` — optional, ignore auto-expires after this date
- Ignored vulnerabilities are shown in a separate section of the report and excluded from `--fail-on` checks
CI Integration
# GitHub Actions example
- name: Security Audit
run: npx audit-mcp-cli --fail-on high# Generic CI
npx audit-mcp-cli --fail-on high && echo "pass" || echo "fail"License
Frequently asked questions
What is audit-mcp-cli?
audit-mcp-cli is audit-mcp-cli
How do I install audit-mcp-cli?
Open the GitHub repository and follow its README. Most MCP servers are added to your client's MCP config, then called by your agent.
Is audit-mcp-cli open source?
Yes — it is hosted on GitHub at https://github.com/double527/audit-mcp-cli and has 1 stars.
Related MCP tools
The Open-Source Multimodal AI Agent Stack: Connecting Cutting-Edge AI Models and Agent Infra
A MCP for Claude Desktop / Claude Code / Windsurf / Cursor to build n8n workflows for you
Browser MCP is a Model Context Provider (MCP) server that allows AI applications to control your browser
A Model Context Protocol (MCP) server and CLI that provides tools for agent use when working on iOS and macOS projects.
The go-to web for your AI coding agent — local-first search, fetch, crawl & research over MCP. No API keys, no cloud, $0/query. Public beta.
MCP Aggregator, Orchestrator, Middleware, Gateway in one docker
Run your own MCP server? See who uses it and what to fix.
Measure it with TrackMCP