MantaGet Started
Docs/Scan Types

8 Security Scanners

Purpose-built scanners covering the OWASP Top 10 for LLM Applications.

OWASP LLM Top 10 Coverage

LLM01 - Prompt Injection
LLM02 - Insecure Output Handling
LLM03 - Training Data Poisoning
LLM04 - Model DoS
LLM05 - Supply Chain Vulnerabilities
LLM06 - Sensitive Info Disclosure
LLM07 - Insecure Plugin Design
LLM08 - Excessive Agency
LLM09 - Overreliance
LLM10 - Model Theft

✓ Full coverage • ◐ Partial coverage (detection only)

MCP Server Audit

Code

Deeply inspect your MCP server for dangerous tool exposure, overprivileged permissions, insecure configurations, and attack surfaces that could let a bad actor hijack your agent's actions.

OWASP Coverage

LLM01 - Prompt Injection vectors
LLM07 - Insecure Plugin Design
LLM08 - Excessive Agency

What We Check

critical
Unrestricted shell execution

Detects shell_exec, eval, system calls without proper sandboxing

critical
Path traversal risks

Identifies unrestricted file read/write operations

high
Tool shadowing

Detects tools that override built-in functions

high
Missing input validation

Identifies parameters without type checking or sanitization

medium
Excessive permissions

Flags tools with overly broad access patterns

medium
Hardcoded secrets

Scans for API keys, tokens, and credentials in code

low
Missing rate limiting

Identifies tools without usage throttling

Prompt Injection Tester

URL

Fire 25+ real-world jailbreak payloads at your LLM endpoint — including DAN attacks, role-play exploits, encoding bypasses, and instruction overrides — to see what your model will do when an attacker gets creative.

OWASP Coverage

LLM01 - Prompt Injection
LLM02 - Insecure Output Handling

What We Check

critical
Direct injection

Payloads that directly manipulate system prompts

critical
DAN jailbreaks

Do Anything Now and other persona-based bypasses

high
System prompt extraction

Attempts to leak system prompts or instructions

high
Role-play exploits

Hypothetical scenarios that bypass safety

medium
Encoding bypasses

Base64, rot13, and other encoding tricks

medium
Context manipulation

False memory and multi-turn attacks

Tool Fuzzer

JSON

Hammer your tool definitions with path traversals, SQL injections, oversized inputs, null values, and malformed JSON to expose how your tools behave under adversarial conditions.

OWASP Coverage

LLM07 - Insecure Plugin Design
Input validation vulnerabilities

What We Check

critical
Path traversal

../../../etc/passwd and similar payloads

critical
SQL injection

SQL payloads in string parameters

critical
Prototype pollution

__proto__ and constructor injection

high
Type confusion

Passing wrong types (string vs number vs array)

high
Boundary testing

Empty strings, huge inputs, negative numbers

high
Null byte injection

Embedded nulls to truncate strings

medium
Malformed JSON

Deeply nested objects, circular references

SSRF Detection

URL

Test any endpoint that fetches external URLs for Server-Side Request Forgery vulnerabilities — catching cases where attackers could redirect your server to hit internal infrastructure, cloud metadata endpoints, or private networks.

OWASP Coverage

LLM07 - Insecure Plugin Design
SSRF vulnerabilities

What We Check

critical
Cloud metadata access

AWS/GCP/Azure metadata URLs (169.254.169.254)

critical
Internal network access

Private IP ranges (10.x, 172.16.x, 192.168.x)

high
Localhost bypass

127.0.0.1, 0.0.0.0, [::1] and other variants

high
DNS rebinding

Time-based DNS resolution attacks

medium
Protocol smuggling

file://, gopher://, dict:// protocols

Data Leak Scanner

Code

Scan your codebase for hardcoded API keys, secrets, tokens, and exposed PII before they end up in a GitHub repo or a production log that someone else is reading.

OWASP Coverage

LLM06 - Sensitive Information Disclosure
Secret management

What We Check

critical
Hardcoded API keys

AWS, OpenAI, Stripe, and other API keys in code

critical
Embedded passwords

Passwords and tokens in plaintext

high
PII exposure

Names, emails, SSNs, credit cards in code

critical
Private keys

RSA, SSH, and other private keys

high
Database credentials

Connection strings with embedded passwords

high
JWT secrets

Hardcoded JWT signing secrets

Permission Analyzer

JSON

Audit your tool and agent permission configurations for privilege escalation paths, overly broad scopes, and access patterns that violate least-privilege principles.

OWASP Coverage

LLM08 - Excessive Agency
Principle of least privilege

What We Check

high
Unrestricted file access

Tools that can read/write any path

high
Unrestricted network

Outbound connections without restrictions

critical
System command access

Shell execution without sandboxing

high
Database access

Direct DB connections without query restrictions

critical
Privilege escalation

Paths to gain elevated permissions

medium
Missing authentication

Tools that don't verify caller identity

Output Validator

Code

Analyze code that renders or displays LLM-generated output for XSS vulnerabilities, HTML injection, and unsafe rendering patterns that could turn your chatbot into an attack vector.

OWASP Coverage

LLM02 - Insecure Output Handling
XSS prevention

What We Check

critical
XSS vulnerabilities

Unsafe innerHTML, dangerouslySetInnerHTML

high
HTML injection

Unescaped HTML in LLM output

critical
Script injection

JavaScript execution in rendered content

medium
Link injection

Malicious URLs in generated links

medium
CSS injection

Style-based attacks in output

medium
Markdown injection

Unsafe markdown rendering

Dependency Scanner

JSON

Check every AI and LLM package in your stack against known CVE databases to surface vulnerable dependencies before they become your weakest link.

OWASP Coverage

LLM05 - Supply Chain Vulnerabilities

What We Check

varies
Known CVEs

Checks npm/pip packages against vulnerability databases

critical
Malicious packages

Known typosquatting or compromised packages

low
Outdated packages

Packages significantly behind latest version

high
AI/LLM library vulns

LangChain, OpenAI SDK, and other AI packages

info
License compliance

Identifies copyleft and problematic licenses

Ready to scan?

Run your first security scan in under 5 minutes.

Get Started