Free · no signup
MCP Scanner
Paste a public MCP server URL and see which tools can delete, export, send data out or move money. No tool is ever called.
See servers already scannedCheck a Claude Code configVulnify checks every sensitive action your agent takes, like exporting, deleting or paying, and answers allow, review or block. With the reasons.
Works withClaudeCursorOpenAI Agents SDKLangChainNodePythonMCP
How it works
Runtime authorization in three steps, with no change to how your agent thinks.
Before a sensitive action, one line of code asks Vulnify.
await vulnify.guard('EXPORT_DATA')A score from 0 to 100 and the reasons, in plain words.
Reviews arrive in Slack. Every decision lands in a verifiable audit trail.
Playground
Pick an action and see what the real engine decides.
Pick what the agent wants to do
The agent mails 12,000 customer rows to a personal Gmail “for backup”.
EXPORT · CUSTOMER_PII · 12,000 records · → growth.team.backup@gmail.com
Real engine, demo policies. Nothing you type is stored.
Product
Everything your team needs to trust what agents do.
Free resources
Free · no signup
Paste a public MCP server URL and see which tools can delete, export, send data out or move money. No tool is ever called.
See servers already scannedCheck a Claude Code config
Research
How many tools can send, delete or move money, and how far their safety hints can be trusted.
Templates · CC0
10 ready-made policies for personal, health and financial data. Open one in Vulnify in monitor mode.
Browse templatesFor developers
Node and Python SDKs, LangChain, or the MCP server in Claude and Cursor. Policies live in git as YAML.
import { Vulnify } from '@vulnify/sdk';
const vulnify = new Vulnify({ apiKey: process.env.VULNIFY_API_KEY });
await vulnify.guard(
{ agent: 'sales-copilot', action: 'EXPORT_DATA', recordsAffected: 48120 },
() => exportContacts(),
);
npm install @vulnify/sdkPricing
No annual plan and no overage charge.
Prices in USD. Brazilian accounts pay in BRL.
FAQ
Free to start. Running in about two minutes.