pkgverdict
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@pkgverdictcheck if unused-imports is safe to install, or if it's a typosquat"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
pkgverdict
Check an npm package before you install it. One command, one verdict: GO, RISKY, FAKE, or MISSING — in plain words your AI coding agent can read too.
npx -y pkgverdict check react-markdown
# [GO] react-markdown@10.1.0
# - established, 38.4M/wk, no install scripts, no advisories, no lookalike name.
npx -y pkgverdict check unused-imports
# [FAKE] unused-imports@0.0.1-security
# - Flagged malicious: MAL-2025-48781. Do not install.
# - Name rides on eslint-plugin-unused-imports (suffix of popular name) and the
# package is barely downloaded — the classic typosquat shape.
npx -y pkgverdict check react-super-hooks-plus
# [MISSING] react-super-hooks-plus
# - react-super-hooks-plus is not on the npm registry. If an AI assistant
# suggested it, the name was hallucinated.Why
AI assistants now pick half the packages developers install — and they hallucinate names attackers have already registered (slopsquatting), while typosquats and hijacked releases keep shipping postinstall malware. npm audit only sees published advisories, after the fact. pkgverdict checks before the install: existence, lookalike names, package age, install scripts, maintainers, downloads, and OSV advisories.
Related MCP server: pkg-oracle
Use it inside your AI coding tool
One command sets it up. In your project directory, run:
npx -y pkgverdict init --agent claude-codeThat installs the MCP server config plus the instruction snippet teaching your agent the one behavior rule: check pkgverdict before installing. Supported tools: claude-code, cursor, copilot (VS Code), windsurf, cline, and generic (init --list shows them all; --all sets up everything; --dry-run previews without writing). Restart your tool, verify with a "list your MCP tools" prompt, and the pkgverdict server should appear with 2 tools (check_package, scan_dependencies). The underlying pack lives in integrations/ if you prefer to install by hand.
CLI
# Verdict on one package (name or name@version)
npx -y pkgverdict check lodash
# Check every dependency in package.json, worst first
npx -y pkgverdict scan
npx -y pkgverdict scan ./my-app --no-dev
# Check your setup (Node version, registry + OSV reachability, cache)
npx -y pkgverdict doctor
# JSON output for scripting and CI
npx -y pkgverdict check left-pad --jsonExit codes: 0 when everything is GO (or RISKY without --strict), 1 when a FAKE or MISSING package is found, 2 on usage errors. For CI gates that should fail on anything suspicious: pkgverdict scan --strict.
Verdicts
Verdict | Meaning |
GO | Established package, no red flags. Install away. |
RISKY | Caution: install scripts, very young, obscure, deprecated, no repo/license, known advisories, or a lookalike name on an otherwise fine package. Read before installing. |
FAKE | Looks actively bad: typosquat-shaped name on a new/obscure package, hostile install script, or a malicious (MAL-) advisory. Do not install. |
MISSING | Not on the registry at all — a typo or a hallucinated name, with a "did you mean" suggestion when one fits. |
Signals come from the npm registry (metadata, downloads), the OSV database (advisories), and a bundled list of well-known packages for lookalike detection. Responses cache under ~/.cache/pkgverdict (6h metadata, 1h advisories); delete that directory to force fresh checks.
Run it with Docker
pkgverdict also ships as a container image on GHCR: ghcr.io/cloudcorprecords/pkgverdict.
Use it as a stdio MCP server:
docker run -i ghcr.io/cloudcorprecords/pkgverdictAny CLI command can be passed after the image name instead of the default mcp argument:
docker run ghcr.io/cloudcorprecords/pkgverdict <command> [args]MCP server
Run it with npx -y pkgverdict mcp (stdio), or point your agent at the pkgverdict command with the single argument mcp. Two tools: check_package (verdict on one package) and scan_dependencies (worst-first verdicts for a project's package.json).
Development
npm install
npm run build
npm test # offline fixture corpus: every verdict rule pinned to a caseLicense
MIT.
This server cannot be deployed
Maintenance
Related MCP Connectors
Check npm packages for typosquats, hallucinated names, install scripts and risky new releases.
Verify npm packages before your AI agent installs them: hallucinations, advisories, API drift.
Detect malicious or vulnerable npm packages: registry search, OSV.dev and GitHub advisory lookups
Check an npm/PyPI package before an AI agent installs it: 5 focused supply-chain tools.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceAudits npm packages for supply-chain attacks (typosquatting, malicious install scripts, credential exfiltration) before installation, returning a SAFE/SUSPICIOUS/DANGEROUS verdict.MIT
- AlicenseNot gradedqualityBmaintenanceAn MCP server that verifies npm and PyPI packages before installation, checking for existence, known vulnerabilities, OpenSSF scorecard, and typosquatting, returning an ALLOW/WARN/BLOCK verdict.MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI coding agents to automatically verify npm packages against the live registry before installation, flagging hallucinated, slopsquatted, or otherwise suspicious packages with risk verdicts.1 npmMIT
- AlicenseAqualityAmaintenanceEnables AI coding agents and CI to vet npm dependencies before they reach the lockfile, flagging hallucinated, slopsquatted, or otherwise risky packages with evidence-backed verdicts.361 npm4MIT