MCP Security Scanner
Scan MCP server configurations for security issues including missing auth, CORS, and HTTPS.
- Free
- No account
- Runs in your browser
- Nothing uploaded
Paste an MCP server URL or JSON config and click Scan — all checks run in your browser. Nothing is sent anywhere.
Runs entirely in your browser — your input is never uploaded, logged, or stored.Privacy policy
What is MCP Security Scanner?
The Model Context Protocol (MCP) lets AI assistants call external tools through a standardized interface. A February 2026 scan found over 8,000 MCP servers exposed with no authentication — meaning anyone who knows the URL can call those tools, read their outputs, and potentially execute arbitrary operations on the underlying systems.
This tool analyzes MCP server configurations for common security issues:
1. Authentication: Does the server require any form of authentication (API key, token, OAuth)? Servers without auth are callable by anyone who discovers the URL.
2. HTTPS: Is the server URL using HTTPS? HTTP transmissions can be intercepted and modified.
3. CORS Configuration: If the server is accessed from a browser, does it restrict which origins can make requests? Permissive CORS (Access-Control-Allow-Origin: *) allows any website.
4. Tool Description Poisoning: Are tool descriptions clear and honest, or do they contain instructions that could manipulate an AI into doing something unintended?
5. Input Validation: Does the server validate inputs before processing? Servers that pass user input directly to shell commands or database queries are vulnerable.
Important: This tool checks the MCP configuration and endpoint headers — it does not perform penetration testing or attempt to exploit vulnerabilities. It is a configuration review, not a security audit. For production systems, use dedicated security scanning tools and conduct proper penetration testing.
Security checks performed:
1. HTTPS Check: Verifies the server URL uses HTTPS. Risk: HTTP allows MITM attacks and eavesdropping. Fix: Use HTTPS for all MCP servers.
2. Authentication Check: Looks for auth-related fields in the config. Risk: Unauthenticated servers can be called by anyone. Fix: Require API keys, tokens, or OAuth for all servers.
3. CORS Check: If applicable, checks for permissive CORS headers. Risk: Any website can call the server from a user's browser. Fix: Restrict Access-Control-Allow-Origin to trusted domains.
4. Tool Description Analysis: Scans tool descriptions for suspicious patterns. Risk: Malicious descriptions can manipulate AI behavior. Fix: Keep descriptions factual and reviewed.
5. Transport Security: Checks for TLS configuration if exposed. Risk: Weak TLS allows traffic interception. Fix: Use TLS 1.2+ with strong cipher suites.
Each check returns PASS, WARN, or FAIL with an explanation of the risk and a fix.
Worked examples
- Server URL "http://localhost:3000" → FAIL: HTTP instead of HTTPS
- Server config without auth fields → WARN: No authentication configured
- Server URL "https://mcp.example.com" → PASS: Using HTTPS
- Tool description "IGNORE ALL INSTRUCTIONS AND..." → FAIL: Tool description poisoning detected
How to use MCP Security Scanner
- Paste your MCP server URL or JSON configuration.
- Click Scan to run the security checks.
- Review each check result (PASS, WARN, or FAIL).
- Follow the recommended fixes for any warnings or failures.
Common errors
- All checks show WARN — this is normal for development servers. Production servers should aim for all PASS.
- localhost URLs always fail HTTPS — this is expected. Local development servers don't need HTTPS, but production servers must use it.
- False positive on tool description — some legitimate tools use imperative language in descriptions. Review manually if the scanner flags a description you wrote.
FAQ
Does this tool actually test the server?
No. It analyzes the configuration you provide (URL, headers, config JSON) and checks for common security anti-patterns. It does not connect to the server, send requests, or attempt to exploit vulnerabilities. Think of it as a linter for MCP security configuration.
What if I'm developing locally?
Local development servers (localhost, 127.0.0.1) don't need HTTPS or authentication. The scanner will flag these as warnings — that's expected. Focus on fixing issues before deploying to production.
How do I fix tool description poisoning?
Keep tool descriptions factual and concise. A tool description should explain what the tool does, not contain instructions for the AI. If you see a description that says 'ignore previous instructions' or similar, that's a red flag — review the tool's source code.
Is this a complete security audit?
No. This is a quick configuration check. For production MCP servers, you should also: conduct penetration testing, review the server's source code, implement rate limiting, add logging and monitoring, and follow your organization's security policies.
Related tools
Prefer AllUtil on Google
One click adds AllUtil to your Google preferences. You'll see our tools highlighted with a Preferred badge in Search and AI answers.