Security for AI-written code

Your AI installs things.
Some of them don't exist.

ghostcheck reads what your coding agent wrote before anything installs or ships: invented packages, look-alikes, vulnerable versions, leaked secrets and risky install scripts.

agent · npm installready
Scanned 0 / 0Ghosts blocked 0Local · no upload
Pro waitlist

Let your agent move fast without shipping a ghost.

What it catches

Six quiet ways AI code goes wrong.

01

Invented packages

Names the model made up. Attackers register them. Checked against the registry before install.

express-rate-guard
02

Look-alike names

One letter off a popular package, flagged with the one you meant.

reqeusts → requests
03

Vulnerable versions

Pins with published advisories, and the first safe version.

jsonwebtoken 8.5.0 → 9.0.0
04

Leaked secrets

Keys and tokens pasted into code or config, caught before the commit.

STRIPE_KEY = "sk_live_…"
05

Risky install scripts

Install hooks that download, execute or reach into your home folder.

postinstall: curl … | sh
06

Agent command guard

Pro checks the shell commands your agent wants to run, before they run.

rm -rf ~/
Where it runs

One engine. Every place you code.

ExtensionVS Code
Open VSXCursor
PluginClaude Code
MCP toolCodex
CI + pre-commitCLI
Pricing

Pay once. Keep it for life.

Free
$0
  • Invented + look-alike package checks
  • Daily scan limit
  • Editor + CLI
Pro · waitlist price
$9lifetime
  • Unlimited scans, every check
  • Vulnerable versions, secrets, install scripts
  • Agent command guard for Claude Code + Codex
  • CI mode and every future update
Launching soon

Know what your AI installs.