Claude MCP Plugin

agentverif for Claude

Connect Claude to the agent certification registry. Claude verifies agents automatically — before you execute them.

Add to Claude → View on GitHub →

https://mcp.agentverif.com

How it works

Automatic. Instant. Free.

One connection. Claude handles everything from there.

SCAN OWASP LLM Top 10

SIGN License + hash

VERIFY Registry check

Automatic verification

Claude calls verify_agent whenever you mention running, installing, or trusting an AI agent. No manual steps.

5 clear verdicts

VERIFIED, UNREGISTERED, UNSIGNED, TAMPERED, REVOKED. No ambiguity. You always know exactly what you’re running.

Zero configuration

No account. No API key. No subscription. Free public service. One URL to connect.

Setup

Connect in 30 seconds

Choose your Claude client. No authentication required.

Claude.ai

Go to claude.ai → Settings → Connectors
Click "Add connector"
Enter: https://mcp.agentverif.com
Click Connect — done

Open Claude Settings →

Claude Desktop

~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "agentverif": {
      "url": "https://mcp.agentverif.com"
    }
  }
}

Save the file and restart Claude Desktop.

Examples

See it in action

Real Claude conversations with agentverif connected.

Verify before executing

"I want to run this agent, license AC-84F2-91AB. Is it safe?"

✅ VERIFIED — agentverif certified
License: AC-84F2-91AB | Tier: pro
Issuer: agentverif.com
Hash: sha256:abc123... ✓ intact
Safe to execute.
🔗 verify.agentverif.com/?id=AC-84F2-91AB

Catch a tampered agent

"Can you verify this agent: AC-12F3-88AC"

🔴 TAMPERED — Certificate invalid
License: AC-12F3-88AC | Status: HASH MISMATCH
This agent was modified after signing.
Do NOT execute.
🔗 verify.agentverif.com/?id=AC-12F3-88AC

Unsigned agent warning

"Run this agent for me"

🚫 UNSIGNED — No certificate found
No agentverif certificate found.
Unsigned agents carry unknown risk.

Ask the vendor to certify free:
agentverif.com

Scan

Scan before you sign

Claude calls scan_agent to check for OWASP LLM Top 10 violations before any certificate is issued.

Scan refused — score below 70

"Scan this agent: https://example.com/agent.zip"

## Scan Result: Score 45/100 — REFUSED 🔴

LLM01 — Prompt injection via user_input (critical)
Sanitise all user-supplied strings before
passing to the LLM context.

LLM06 — Sensitive data in logs (warning)
Remove PII from debug logging.

Score below 70 — signing refused.

Scan passed — ready to sign

"Scan this agent: https://example.com/safe-agent.zip"

## Scan Result: Score 88/100 — PASS ✅

No critical or warning violations found.

Ready to sign at sign.agentverif.com

Reference

Verification statuses

Every possible result, explained.

Status	Meaning	Action
✅ VERIFIED	Signed, registered in public registry, hash intact	Safe to execute
⚠️ UNREGISTERED	Signed locally, not in public registry	Use with caution
🚫 UNSIGNED	No certificate found	Ask vendor to certify
🔴 TAMPERED	Modified after signing — hash mismatch	Do NOT execute
🔴 REVOKED	Licence cancelled by issuer	Do NOT execute

Tools

Available MCP tools

Both tools are available as soon as you connect.

Tool	When Claude calls it	What it returns
scan_agent	You ask to scan an agent ZIP URL	Score 0–100, OWASP violations with explanations and fixes
verify_agent	You mention running, installing, or trusting an AI agent	VERIFIED / UNREGISTERED / UNSIGNED / TAMPERED / REVOKED + hash check

Sell AI agents?

Get certified free. Buyers trust verified packages. Unsigned agents get ignored.

Certify your agent free →