Skip to main content

Vulkro Labs · free tools

Check what your AI assistant adds, before it lands.

Thirteen free, keyless commands that vet packages that do not exist or are malicious, MCP servers, tools, skills and memory files. Your source code never leaves your machine; some checks look up public package metadata.

Install the free toolsThe Labs docs

Free · keyless · your code stays on your machine

terminalvulkro-live

$ vulkro-live verify express lodahs axios-retry-helper

  • npmexpressOKlatest 5.2.1, 109089346 weekly downloads
  • npmlodahsMALICIOUSflagged malicious by OSV (MAL-2025-25502): Malicious code in lodahs (npm)
  • npmaxios-retry-helperMISSINGnot found in the registry

3 checked: 1 OK, 1 MALICIOUS, 0 LOOKALIKE, 0 VULNERABLE, 1 MISSING, 0 SUSPICIOUS

The example run from the verify docs: one real package, one malicious lookalike, one name that does not exist.

01How it works

One download. No account, no key.

Install it once and run any check before you accept what your assistant suggests.

Keyless
No account, no key, no sign-in. Download it and run it.
On your machine
Your source code is never uploaded. The checks that need public package data look up package names and versions in the public registries and advisory feeds, and nothing else.
Free, and separate
The free tools vet what enters your project. Vulkro, the licensed scanner, analyses the code you write.
terminal
$ curl -fsSL https://dist.vulkro.com/install-live.sh | bash
# no account
# no key
# your source code stays on the machine

02What it checks

Thirteen checks, one binary.

Everything an assistant can pull into a project, checked before it lands. Every check can also write SARIF, so results land where your other security tools report.

Every command, in the docs

Packages

Before anything is installed.

  • verifyIs this package real, and is it safe? Invented names, typosquats, known-malicious releases and known vulnerabilities.
  • foreseeThe names an assistant is likely to invent for your stack, so they can be blocked before they are suggested.
  • provenanceDoes this package carry published build provenance? A presence check only: the attestation is not verified.
  • trustdbClear a package or server once and every check trusts that version until it changes, in a file you can commit.

MCP servers and tools

Before your assistant trusts them.

  • inspectIs this MCP server safe to add? Checks the package behind it and answers green, review or avoid.
  • wardenTool descriptions and results checked for prompt injection, tool poisoning, hidden characters and exfiltration.
  • lock and driftFingerprint a server's tools, then see what changed field by field, so a silent change is reported.
  • cardcheckAn agent card checked for the permissions and endpoints it claims, against what it can actually reach.

Your assistant setup

What it already runs on.

  • auditEvery configured MCP server, plus the rules, skills, instruction files and network-reaching hooks.
  • skillscanWhat your skills, slash commands and subagents actually run, not just their prose, with stealer behaviour flagged.
  • memcheckMemory files checked for planted instructions and credentials that should never have been written down.
  • mcpThe checks as an MCP server, so your assistant vets its own suggestions in the loop.

Check new packages and MCP servers before you add them.

Install the free toolsVulkro Core