About Osv Ui Mcp
osv-ui scans dependency manifests against OSV.dev and presents vulnerabilities in a local browser dashboard. It is for developers reviewing dependency risk across projects, with fixes applied only after explicit confirmation.
Key features
- 01Scan manifests across seven ecosystems
- 02Show vulnerabilities and risk scores
- 03Display upgrade guidance and fix commands
- 04Export reports and SBOMs
- 05Compare scans against a baseline
- 06Review fixes before applying
Use cases
- 01Review dependency vulnerabilities across a monorepo
- 02Generate CI reports and compare pull request scans
- 03Inspect findings in a visual dashboard
MCP support
osv-ui also runs as an MCP server for AI agents. It can scan projects, open the dashboard for human review, and apply fixes after explicit confirmation.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
4Saved tool list · updated 3w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| path | string | No | Absolute or relative path to the project directory. Defaults to current working directory. |
| severity_filter | enum | No | Only return vulnerabilities at this severity or above. Default: all.all | critical | high | moderate | low |
| offline | boolean | No | If true, skip OSV.dev query and only parse manifests. Default: false. |
Details
From the project README
osv-ui
A beautiful, zero-config visual CVE dashboard for npm, Python, Go, Rust, Java, PHP, and Ruby projects.
One command. No signup. No API key. Runs 100% locally — your code never leaves your machine.
🇻🇳 Tiếng Việt · 🇺🇸 English · 🇨🇳 中文 · 🇯🇵 日本語
The problem
$ npm audit
# ... 300 lines of this ...
# moderate Regular Expression Denial of Service in semver
# package semver
# patched in >=7.5.2
# ...
# 12 vulnerabilities (3 moderate, 6 high, 3 critical)
Nobody reads that. Security gets ignored. Dependencies stay vulnerable.
The solution
npx osv-ui
→ Opens a dashboard. Every CVE, every fix, all your services. Done.
Why give it a try?
- Zero-config: No complex setup, no signup, no API key required.
- Privacy First: Analysis is done 100% on your machine.
- Fast & Visual: Real-time Risk Scores, vulnerability charts, and clear upgrade guides in seconds.
- Multi-platform: Native support for Node.js (npm), Python, Go, Rust, Java, PHP, and Ruby.
Features
| 🌐 Multi-Ecosystem | Scans package-lock.json, pnpm-lock.yaml, yarn.lock, Pipfile.lock, poetry.lock, requirements.txt, go.sum, Cargo.lock, pom.xml, composer.lock, Gemfile.lock |
| 📡 Live CVE data | Powered by OSV.dev — updated daily from NVD, GitHub Advisory, PyPI Advisory. No API key. |
| 🏢 Multi-service | Scan your entire monorepo in one command — frontend, backend, workers, ML services |
| 💊 Fix guide | Dependabot-style upgrade table: current version → safe version + one-click copy command |
| 🔌 Built-in REST API | Power your own security dashboards with GET /api/data or CLI export flags |
| 🎯 Risk score | 0–100 per service so you know where to focus first |
| 🔍 CVE drill-down | Click any row — CVSS score, description, NVD link, GitHub Advisory link |
| 🌙 Dark Mode | Eye-friendly security audits, day or night |
Quick start
Scan current directory:
npx osv-ui
Scan a monorepo (multiple services at once):
npx osv-ui ./frontend ./api ./worker ./ml-service
Auto-discover all services under the current directory:
npx osv-ui -d
Add to your package.json scripts:
{
"scripts": {
"audit:ui": "npx osv-ui",
"audit:all": "npx osv-ui ./frontend ./api ./worker"
}
}
--discover, -d Auto-find service dirs that contain a supported manifest
--port=2003 Use a custom port (default: 2003)
--json[=file] Save report as JSON without opening browser (defaults to osv-report.json)
--html[=file] Save report as HTML without opening browser (defaults to osv-report.html)
--cyclonedx[=file] Save CycloneDX SBOM JSON (defaults to osv-sbom.cdx.json)
--spdx[=file] Save SPDX SBOM JSON (defaults to osv-sbom.spdx.json)
--baseline=file Compare with a previous --json report
--markdown[=file] Save a Markdown PR/comment report (defaults to osv-report.md)
--fail-on=level Exit non-zero for critical/high/moderate/low findings
--webhook-url=url POST matching findings to a webhook
--webhook-severity=level Webhook threshold (default: critical)
--watch Keep dashboard running and re-scan when manifests change
--no-open Don't auto-open the browser
--offline Skip OSV.dev lookup — parse manifests only
-h, --help Show help message
🤖 AI Agent Integration (MCP)
osv-ui is now a Model Context Protocol (MCP) server. This allows AI agents like Claude Desktop, Cursor, and Claude Code to:
- Scan your project for CVEs automatically.
- Open the visual dashboard for you to review findings (Human-in-the-loop).
- Apply fixes after your explicit confirmation.
Quick setup (npx):
{
"mcpServers": {
"osv-ui": {
"command": "npx",
"args": ["-y", "osv-ui-mcp"]
}
}
}
See the MCP Package README for detailed setup instructions.
🔌 Powerful built-in API
osv-ui isn't just a dashboard; it's a security data engine.
Once the dashboard is running, you can pull the raw security data for your whole project:
# Get full JSON payload for all services
curl http://localhost:2003/api/data
# Use it in your custom scripts
curl -s http://localhost:2003/api/data | jq '.[0].vulns'
CI reports, PR diffs, and SBOMs
Generate machine-readable reports without opening the browser:
npx osv-ui -d --json=osv-report.json --markdown=osv-report.md --cyclonedx=sbom.cdx.json --spdx=sbom.spdx.json --no-open
Compare a PR scan against a baseline report and fail on newly introduced high+ findings:
npx osv-ui -d --json=current.json --baseline=main-osv-report.json --markdown=osv-pr.md --fail-on=high --no-open
Minimal GitHub Actions flow:
name: osv-ui
on: [pull_request]
jobs:
scan:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/setup-node@v4
with:
node-version: 24
- run: npx osv-ui -d --json=current.json --markdown=osv-pr.md --cyclonedx=sbom.cdx.json --fail-on=high --no-open
- uses: actions/upload-artifact@v4
if: always()
with:
name: osv-ui-report
path: |
current.json
osv-pr.md
sbom.cdx.json
Send new critical findings to a webhook:
npx osv-ui -d --baseline=main-osv-report.json --webhook-url="$SECURITY_WEBHOOK_URL" --webhook-severity=critical --json=current.json
Supported manifest files
| Ecosystem | Files |
|---|---|
| npm / JS | package-lock.json · pnpm-lock.yaml · yarn.lock |
| Python | requirements.txt · Pipfile.lock · poetry.lock · pyproject.toml · uv.lock |
| Go | go.sum |
| Rust | Cargo.lock |
| Java | pom.xml (Maven) |
| PHP | composer.json · composer.lock |
| Ruby | Gemfile · Gemfile.lock |
More ecosystems coming — see Roadmap.
How it works
Your project files
│
├─ package-lock.json ──┐
├─ Pipfile / poetry ──┤──► parser ──► package list
├─ go.sum / Cargo.lock ──┘
│
▼
OSV.dev batch API (free, no key)
│
▼
CVE matches + fix versions
│
▼
Express server → browser dashboard
http://localhost:2003
CVE data comes from OSV.dev — a free, open database maintained by Google that aggregates:
- 🇺🇸 NVD — NIST National Vulnerability Database
- 🐙 GitHub Advisory Database (GHSA)
- 🐍 PyPI Advisory Database
- 📦 npm Advisory Database
- 🦀 RustSec · Go Vuln DB · OSS-Fuzz · and more
Updated daily. No account. No rate limit. No vendor lock-in.
Works great alongside osv-scanner (Google)
osv-ui and osv-scanner use the same OSV.dev data source. osv-ui adds the visual layer that osv-scanner lacks:
- Browser dashboard instead of terminal output
- Multi-service sidebar
- Dependabot-style upgrade guide with copy commands
vs alternatives
| osv-ui | npm audit | Snyk | Dependabot | |
|---|---|---|---|---|
| Visual dashboard | ✅ | ❌ terminal only | ✅ | ✅ |
| npm support | ✅ | ✅ | ✅ | ✅ |
| Python support | ✅ | ❌ | ✅ | ✅ |
| Multi-service in one view | ✅ | ❌ | ✅ paid | ✅ |
| No signup required | ✅ | ✅ | ❌ | ❌ |
| Works on GitLab Free | ✅ | ✅ | ❌ | ❌ |
| Self-hosted / local | ✅ | ✅ | ❌ | ❌ |
| Fix commands | ✅ | partial | ✅ | ✅ |
| Open source | ✅ | ✅ | ❌ | ❌ |
GitLab CI — block deploys on critical CVEs
No Dependabot on GitLab Free? Add this to .gitlab-ci.yml:
audit:
stage: test
image: node:24-alpine
script:
- npm audit --json > /tmp/audit.json || true
- |
node -e "
const r = require('/tmp/audit.json');
const crit = Object.values(r.vulnerabilities || {})
.filter(v => v.severity === 'critical').length;
if (crit > 0) {
console.error('BLOCKED: ' + crit + ' critical CVE(s). Run: npx osv-ui');
process.exit(1);
}
console.log('OK: no critical vulnerabilities');
"
artifacts:
paths: [/tmp/audit.json]
when: always
Requirements
- Node.js >= 22
- Internet access for OSV.dev queries — or use
--offline - npm projects: run
npm installfirst sopackage-lock.jsonexists - Python projects: any of the supported manifest files listed above
Roadmap
All contributions are welcome. If you want to work on something, open an issue first so we can coordinate.
- Go support — parse
go.sum/go.mod - Rust support — parse
Cargo.lock - Java / Maven — parse
pom.xml - PHP / Composer — parse
composer.lock - Ruby / Bundler — parse
Gemfile.lock - Export report — save as HTML / JSON
- Dark mode — eye-friendly dashboard UI
- GitHub Actions / CI diff — generate Markdown PR comments and fail on new CVEs
- SBOM export — CycloneDX / SPDX format
- Watch mode — re-scan on manifest file changes
- Slack / webhook — notify on new critical CVEs
- Parser hardening — Maven property inheritance, lockfile edge cases, workspace layouts
- Live dashboard refresh — push watch-mode updates to an open browser tab without reload
Contributing
This project is built by the community. All skill levels welcome.
Good first issues:
- Write unit tests for the parsers
- Improve Python parser edge cases
- Improve Maven/Gradle and workspace parser edge cases
# Clone and run locally
git clone https://github.com/toan203/osv-ui
cd osv-ui
npm install
# Run against your own project
node bin/cli.js /path/to/your/project
# Run against multiple services
node bin/cli.js ./frontend ./backend
Please read CONTRIBUTING.md for code style and PR process.
License
MIT — use it, fork it, embed it, build on it. Attribution appreciated but not required.
Did osv-ui catch a real CVE in your project?
A ⭐ helps other developers find this tool.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.