An Unreviewed MCP Server Can Expose More Than You Expect

Each MCP server you connect gives agents new tools, and each tool can read data, call APIs, or take actions on your behalf. In two months, Enkrypt AI scanned 268,000+ tools across 25,000 MCP servers and found 143,000+ vulnerabilities affecting 73% of those servers. If you can't see what a server exposes, you can't decide whether to trust it. What the scan checks:

Injection Risk

Tool inputs that malicious prompts or retrieved content can manipulate, including indirect prompt injection (hidden instructions in content the agent retrieves).

Sensitive Data Risk

Paths where personally identifiable information (PII) can leak through tool inputs, outputs, and connectors.

Response Integrity Risk

Tool responses that can include hidden instructions the agent then follows.

Policy Risk

Tools and workflows that break your business rules or skip required approvals.

A Report Your Security Team Can Act On

Each MCP server you connect gives your agents new tools, and each tool brings its own permissions, data access, and ways to fail. MCP Scanner inventories what's connected, maps what each tool can reach, and checks for four kinds of risk.

MCP Inventory

Tools, capabilities, and permissions in one list.

Risk Findings

Prioritized issues with severity, proof, and impact.

Remediation Plan

Least-privilege changes, allowlists, and approval gates to apply next.

Audit Evidence

Exportable summaries ready for security reviews and audit workflows.

Try Enkrypt AI MCP Scanner for Free

Get a security scan report in as little as 10 minutes.

1. Submit Your Server

Enter your email and the server’s GitHub repo, npm package, or remote endpoint.

2. Automated Scan

MCP Scanner discovers tools, analyzes capabilities and permissions, and surfaces high-risk exposure.

3. Get Your Report

When the scan finishes, we email you a link to your report. One free report per email.

Prefer to automate? Start scans from code with the Enkrypt AI Python software development kit (SDK), or use the MCP Hub API for GitHub repos, npm packages, and remote endpoints. Read the MCP Hub API Docs

Found Risk? Enforce Policy in Production.

One scan is a start. Secure every AI agent you ship. Enkrypt AI by Anaconda covers the AI you build, connect and run. Get a demo and see what a scan can’t show you on its own.

Frequently Asked Questions

Is the scan free?

Yes. Enter your email and an MCP server URL, and we will send your report at no cost. To get a demo of internal server scans, contact sales.

MCP servers and their tool surfaces: configuration, permissions, tool schemas, endpoints, and common vulnerability patterns.

GitHub repositories, npm packages, and remote MCP endpoints.

Scan a server before you connect it, after you update its tools, and on a regular schedule. Run on-demand scans during security reviews and incident response.

Yes. We can set up private scanning for internal servers. Get a Demo today.

Each finding is scored on exploitability, impact (data access, action execution, and permission scope), and exposure (public or internal).

Yes. Untrusted MCP servers and poisoned tool catalogs are treated as supply chain risk, with allowlist and denylist guidance.

Enkrypt AI MCP Scanner finds risk before deployment. MCP Gateway enforces policy at runtime and blocks unsafe tool calls.