Alternatives to 42Crunch for LLM gateway boundary test

What middleBrick covers

  • Black-box scanning with no agents or SDK integration
  • Under-one-minute scan time with prioritized findings
  • Covers OWASP API Top 10 (2023) LLM gateway risks
  • OpenAPI 3.0/3.1 and Swagger 2.0 parsing with $ref resolution
  • Authenticated scans with header allowlist and domain verification
  • CI/CD integration via GitHub Action and MCP Server support

Purpose and scope for LLM gateway boundary testing

This tool targets the boundary of an LLM gateway and the surrounding API surface. It focuses on what can be discovered through read-only interaction: header handling, parameter parsing, and error paths. The scanner runs black-box checks, requires no agents or code access, and completes in under a minute. It does not perform intrusive exploit steps such as SQL injection or command injection, which are outside its scope.

Detection coverage aligned to standards

Findings map to OWASP API Top 10 (2023), which covers common risks relevant to LLM gateways. The scanner also supports audit evidence for SOC 2 Type II and PCI-DSS 4.0 by surfacing control-related observations. Specific checks include:

  • Authentication issues such as JWT misconfigurations, alg=none, and missing claims.
  • BOLA and IDOR via sequential ID enumeration and adjacent ID probing.
  • BFLA and privilege escalation through admin endpoint probing and role leakage.
  • Input validation checks, including CORS wildcard with credentials and dangerous HTTP methods.
  • SSRF indicators, focusing on URL-accepting parameters and internal IP detection patterns.
  • LLM / AI Security with 18 adversarial probes across Quick, Standard, and Deep tiers, targeting system prompt extraction, instruction override, jailbreak patterns, data exfiltration, token smuggling, and multi-turn manipulation.

OpenAPI and authenticated scanning relevance

The scanner parses OpenAPI 3.0, 3.1, and Swagger 2.0 with recursive $ref resolution, cross-referencing spec definitions against runtime behavior. This helps identify undefined security schemes, deprecated operations, and missing pagination that may affect LLM gateway design.

Authenticated scanning (Starter tier and above) supports Bearer, API key, Basic auth, and Cookie. Domain verification is enforced via DNS TXT record or an HTTP well-known file so only the domain owner can scan with credentials. Forwarded headers are limited to Authorization, X-API-Key, Cookie, and X-Custom-*.

Product capabilities and integration options

The Web Dashboard centralizes scans, report viewing, score trends, and branded compliance PDF downloads. The CLI supports middlebrick scan <url> with JSON or text output. A GitHub Action can gate CI/CD, failing the build when the score drops below your threshold. An MCP Server enables scanning from AI coding assistants such as Claude and Cursor. An API client is available for custom integrations.

Continuous monitoring (Pro tier) provides scheduled rescans every 6 hours, daily, weekly, or monthly, diff detection across scans, and email alerts rate-limited to 1 per hour per API. HMAC-SHA256 signed webhooks disable automatically after 5 consecutive failures.

Limitations and responsible usage

middleBrick is a scanner that detects and reports with remediation guidance; it does not fix, patch, block, or remediate. It does not detect business logic vulnerabilities, which require domain-specific human review. Blind SSRF is out of scope due to the need for out-of-band infrastructure. The tool does not replace a human pentester for high-stakes audits.

Compliance framing uses alignment language for frameworks outside PCI-DSS 4.0, SOC 2 Type II, and OWASP API Top 10 (2023). For example, findings may help you prepare for or support audit evidence relevant to other standards, but the tool is not certified as compliant with any regulation.

Frequently Asked Questions

Can this scanner test LLM-specific endpoints such as chat completions?
Yes. It sends read-only GET and HEAD requests and text-only POST bodies suitable for LLM probes, focusing on boundary behavior and error handling without executing destructive payloads.
Does the tool perform active SQL injection or command injection testing?
No. Those tests require intrusive payloads outside the scanner’s scope and are not performed.
How are LLM jailbreak and prompt injection checks structured?
The scanner runs 18 adversarial probes across three tiers: Quick, Standard, and Deep. These include system prompt extraction attempts, instruction override, DAN and roleplay jailbreaks, data exfiltration, token smuggling, and multi-turn manipulation.
Can authenticated scans be used for CI/CD gates?
Yes. With a Starter tier or higher, you can provide credentials and enforce domain verification. The GitHub Action can fail the build when the score drops below your defined threshold.