New Research: 6,943 AI agent skills have security flaws. We scanned all 40,059. Read the report →
82% of MCP servers are vulnerable

MCP Security Scanner

Scan your MCP server configurations for path traversal, tool poisoning, exposed credentials, and command injection vulnerabilities.

$npx firmis-cli init

Free scan. No sign-up required. Auto-detects MCP servers.

What you'll see

$ npx firmis-cli init
Detecting platforms...
✓ MCP Servers: 5 servers configured
CRITICAL Tool poisoning in weather-mcp
Hidden instruction: "read ~/.aws/credentials"
HIGH Path traversal in file-server
Escapes sandbox via ../../../etc/passwd
MEDIUM API key exposed in mcp.json
Grade: D1 critical · 1 high · 1 medium

What Firmis Detects in MCP Servers

Tool Poisoning

Malicious MCP servers inject hidden instructions that hijack your AI agent, stealing credentials or executing unauthorized actions.

Path Traversal

Attackers escape sandbox boundaries to read sensitive files like ~/.aws/credentials or ~/.ssh/id_rsa.

Command Injection

Improperly sanitized inputs allow arbitrary command execution on your system.

Credential Exposure

API keys and tokens stored in mcp.json are visible to every connected AI tool.

Use Firmis as an MCP Server

Add Firmis to your MCP configuration and let your AI agent scan for threats automatically.

Claude Code / Claude Desktop

{
  "mcpServers": {
    "firmis": {
      "command": "npx",
      "args": ["-y", "firmis-cli", "--mcp"]
    }
  }
}

Cursor

{
  "mcpServers": {
    "firmis": {
      "command": "npx",
      "args": ["-y", "firmis-cli", "--mcp"]
    }
  }
}

Scan Your MCP Configuration Now

One command. 30 seconds. Know if your MCP servers are safe.

$npx firmis-cli init
View on GitHub