Know what you're installing. Scan any MCP server in 60 seconds and get a professional security report with scores, findings, and remediation steps.
Understand what filesystem, network, and shell access the server requests. Know your exposure before you install.
Automated pattern matching for shell injection, SSRF, path traversal, eval usage, hardcoded secrets, and more.
Dependency count, known compromised packages, lockfile presence, and install script detection.
Professional HTML reports scored 0 to 100. Share with your team, attach to pull requests, include in security reviews.
Compare two MCP servers side by side. See which one is safer across every category before you choose.
Add to GitHub Actions in 2 minutes. Automatically scan on every PR and block merges that fail security checks.
Embed an auto-updating security score badge in your README. Show users your server passes mcp-audit.
Full static analysis across all major MCP server languages. Catches eval, shell injection, pickle attacks, unsafe blocks, and more.
Browse all scanned servers with search and filters. Find the safest MCP servers for your project.
Add a rich score card to your docs or website. More than a badge — shows score ring, status, and links to the full report.
Verify ownership of your MCP server. Get email alerts when your security score changes and a verified badge.
Register webhooks via the API to get notified when scans complete. Integrate with Slack, Discord, or your own systems.
Real data from scanning the most popular MCP servers on GitHub.
Data from the leaderboard. Updated weekly. Read the full report.
Whether you're evaluating third-party servers or shipping your own, mcp-audit fits your workflow.
Evaluate every MCP server before it touches production. Get structured findings you can drop into your security review or compliance report.
Scan your server before publishing. Fix findings before users file issues. Display the security badge to build trust with your users.
Running Claude, GPT-4, or Gemini with MCP tools? Know exactly what filesystem, network, and shell access each server requests before it goes live.
Enter the GitHub URL of any public MCP server repository.
The scanner fetches the source, dependencies, and metadata. No AI involved. Deterministic, reproducible analysis.
Receive a scored security report with specific findings and remediation guidance.
Live feed of servers being scanned right now.
200+ MCP servers scanned and ranked. Here are some highlights.
View full leaderboard (200+ servers) →Shipping production AI agents? I’ll review your MCP setup live, map your 3 highest-risk vectors, and show you how to validate each one. No sales pitch.
HN community rate: €19/mo forever — regular price is €29/mo. First 50 subscribers. Expires when we take this down.
No credit card · API key in inbox instantly · 7 days free, then €19/mo
Start scanning for free. Upgrade when you need more.
| Feature | Free | Pro | Team | Manual Audit |
|---|---|---|---|---|
| Scans per day | 1 | Unlimited | Unlimited + API | N/A |
| Findings shown | Top 3 | All | All | All + custom |
| Remediation guidance | -- | Yes | Yes | Expert-written |
| CI/CD integration | -- | API access | API + webhooks | -- |
| PDF / compliance export | -- | Yes | Yes | Board-ready |
| Private repos | -- | -- | -- | Yes |
| OWASP MCP Top 10 | Basic | Full | Full | Full + threat model |
HN founding rate — €19/mo locked for life. Cancel any time.
⚡ Show HN special — loading spots…
Includes onboarding call. Annual billing available (2 months free).
Fixed price. Delivered in 3 business days. No scope creep.
Questions? hello@pyfio.com
Unlimited seats, private scan history, SSO, and custom SLA. Email us for a custom quote.
The automated scanner covers 90% of cases. For the rest — internal servers, compliance requirements, or building MCP tooling from scratch — I do custom work.
⚡ Flash offer: 3 audit slots left · April 8–18 only
Claim audit slot →Reply within 2 hours. Fixed price, no surprises. Delivered in 48h or free.
Your MCP servers change every sprint. A one-time audit has a 30-day shelf life. The Security Program keeps you covered: monthly re-audits, CVE response within 48h, and a compliance trail your security team can hand to auditors.
MCP servers run with significant permissions inside your AI agent setup. A malicious or poorly-written server can exfiltrate data, execute shell commands, or poison your agent's tool context. Scanning before installing takes 60 seconds and can prevent a breach that takes weeks to clean up.
Five categories: (1) Permission scope -- filesystem, network, shell, and environment access. (2) Code safety -- eval(), SSRF, path traversal, hardcoded secrets. (3) Supply chain -- dependency count, lockfiles, known compromised packages, install scripts. (4) Transparency -- LICENSE, README, description quality. (5) Maintenance -- recent commits, open issues, activity signals.
No. The analysis is deterministic static analysis -- pattern matching against known vulnerability signatures. No LLMs involved in the scan. This means results are reproducible and auditable, not probabilistic.
Tool poisoning is when malicious instructions are hidden in MCP tool names or descriptions. When your AI agent reads these descriptions to decide which tool to call, it may execute unintended actions -- data exfiltration, privilege escalation, or overriding user instructions. This is the #1 MCP-specific attack vector.
The free scanner works with public GitHub repositories. For private repos, internal servers, or on-premise deployments, book a manual security audit starting at EUR 2,500. Contact us for details.
Free: 1 scan per day, top 3 findings shown. Pro (EUR 19/mo founding rate): unlimited scans, all findings with remediation, PDF export, API access for CI/CD. Team (EUR 99/mo): up to 10 members, shared dashboard, Slack notifications. Manual Audit: EUR 2,500 one-time for full human review.