Get your MCP server assessed in minutes
Regulatory Signals is an independent technical evaluation provider. This product is not an accredited certification under ISO/IEC 17065 or ISO/IEC 17021. Evaluation findings are technical assessments, not regulatory endorsements or legal opinions.
11 MCP-specific security rules. Signed assessment report. Public registry listing. Enterprise security teams use our allowlist as the source of truth for approved MCP servers.
npm or GitHub URL · Badge for your README
What the MCP Server Audit fixes
Security Vulnerabilities
The problem
An MCP server hands an AI agent real tools, so a flaw in one becomes an attack on the agent.
What we find
Prompt injection hidden in tool descriptions, command injection through exec and spawn, server-side request forgery in fetch tools, secrets leaking through environment read-back, capability escalation, and invisible characters in server source.
Trust & Liability
The problem
Buyers cannot tell who published a server or what it is allowed to touch.
What we find
Unsigned manifests, resource listings with no authentication, and a signed assessment report plus public registry listing you can show a procurement team.
Reliability Failures
The problem
Tools that misbehave under load or on shutdown break the agent that depends on them.
What we find
Missing rate limits and timeouts, race conditions between checking and using a file, and subprocesses left running after the connection closes.
11 MCP vulnerability rules
OWASP LLM Top 10 (2025) aligned · CWE mapped
CWE-1336 · LLM01
CWE-78
CWE-918 · LLM06
CWE-200 · LLM02
CWE-269 · LLM06
CWE-400
CWE-367
CWE-862 · LLM06
CWE-404
CWE-345
CWE-1007 · LLM01
What you receive
Everything included in the $249 one-time audit.
Signed certificate
- HMAC-SHA256 tamper-evident cert JSON
- Trust grade: Gold / Silver / Bronze / None
- Security score 0–100
- Machine-readable /api/mcp-cert/{slug} endpoint
- Badge embed code for README
Audit findings report
- All 11 rules checked against your source
- Per-finding severity: CRITICAL / HIGH / MEDIUM
- CWE and OWASP LLM Top 10 mapping
- Remediation guidance per finding
- Public MCP Trust Registry listing
Pricing
One payment. Results in minutes. No subscription.
Frequently asked questions
What does an MCP server audit cover?
11 rules covering: prompt injection in tool descriptions (mcp-001), command injection via exec/spawn (mcp-002), SSRF in fetch-style tools (mcp-003), secret leak via env-readback (mcp-004), capability escalation (mcp-005), missing rate-limit/timeout (mcp-006), TOCTOU race in file tools (mcp-007), missing auth on resources/list (mcp-008), zombie subprocess on close (mcp-009), unverified manifest signing (mcp-010), and hidden or invisible characters in server source (mcp-011). Mapped to OWASP LLM Top 10 (2025) and CWE.
How long does the MCP audit take?
Source fetch and rule analysis run automatically. The signed certificate JSON and public registry listing are issued automatically when the analysis finishes.
What does the signed certificate include?
A tamper-evident HMAC-SHA256 certificate with: MCP server slug, npm package, version audited, security score (0-100), trust grade (gold/silver/bronze/none), issuance date, and a machine-readable endpoint at /api/mcp-cert/{slug} for IDE badge integration.
Will there be continuous monitoring?
Not yet. Continuous monitoring (auto re-audit on new npm versions, score drift alerts) is not available at the moment. The one-time $249 audit + cert is available now.
Enterprise teams are building allowlists now
Security teams at companies deploying Claude with MCP use our registry to manage approved server lists. Get your server assessed before they build the allowlist without you.
Where to go next
Searchable registry of audited MCP servers, extensions, and AI agents.
Free to browseBrowse RegistryAI-generated codebase audit — Claude, Codex, Gemini, Cursor, Copilot, Lovable.
Free to startRun Free ScanBrowser extension security and privacy audit — permissions and data flows.
$149 one-timeSubmit Extension