AI Security — Static Analysis

Catch what the AI model reads before it ever calls a tool.

MCP Audit is the static analysis engine for MCP Contracts. It runs automatically on every uploaded contract — no live connection, no flag to set — and produces a score, a grade, and a list of findings.

What's real today: content/agent-facing threat detection (prompt injection, tool poisoning, data exfiltration, tool shadowing) runs automatically, server-side, and produces genuine findings with severity and remediation. What's roadmap: structural contract checks — authentication scheme completeness, authorization coverage, tool integrity signing — are design-intent today, not yet scored. We say so below, section by section.
Two kinds of checks

Agent-facing vs. Backend

Every finding is classified by who it can hurt — the AI agent consuming your server, or the server's own security posture.

Live today

Agent-facing — Content Analysis

Three engines analyze the free-text description of every tool and prompt in your contract, automatically, on every audit:

PromptDefenseChecks for absence of defensive language across 12 attack-vector categories.
YARAMatches known malicious signatures — prompt injection, credential harvesting, tool poisoning.
LLM opt-inSemantic reasoning for novel threats signatures can't catch — requires an API key.

Every content finding carries a remediation string with actionable guidance, shown directly in the Audit tab.

Design intent — not yet scored

Backend — Structural Analysis

Eight contract sections have a designed rule set, not yet implemented. Today's score for these sections counts placeholder markers only — not real structural or security analysis.

serverNaming, versioning conventions
integrityMessage signing, rug-pull protection
authenticationScheme completeness, fail-closed defaults
authorizationRole coverage, wildcard-allow rejection
capabilitiesPer-tool security block, risk level, confirmation requirements
throttlingRate limiting, fail-closed on limit exceeded
Compliance mapping

Every finding mapped to the frameworks you already report against.

Findings are tagged against OWASP, the EU AI Act, ISO 42001, NIST AI RMF, and CSA AICM automatically — so a security review doubles as compliance evidence without extra work.

42Crunch MCP compliance tab showing findings mapped to OWASP, EU AI Act, ISO 42001, NIST RMF, and CSA AICM frameworks
5 / 5 frameworks mapped automatically
How it fits together

Audit is the first gate. Scan validates the live server.

Upload contract | v MCP Audit ← content analysis runs here automatically | score + findings (Agent-facing / Backend) | v SQG gate → CI/CD pass / block (single score threshold, default 75 / grade B) | v MCP Scan (live protocol probing + fuzzing against the running server) | v MCP Protect (contract enforced at runtime)

See MCP Scan for what runs against the live server, and MCP Protect for runtime enforcement.

Upload a contract. See what an AI agent would actually read.