Ask your servers if they're
under attack
Connect Claude, Cursor, Windsurf, or any MCP-compatible AI assistant to your Defensia dashboard. Query server status, active threats, banned IPs, and vulnerabilities — in natural language.
Your server web-prod-01 is receiving a brute force SSH attack from 3 IPs in China.
47 events in the last 15 minutes. All 3 IPs have been automatically banned (escalation: 24h ban).
CPU peaked at 23% during the attack but returned to baseline (8%). No action needed.
Your other 3 servers are all clear — no incidents in the last 12 hours.
In the last hour, 72% of requests to smartsalus are from bots: AhrefsBot (340 req), SemrushBot (218 req), and an unknown crawler (156 req). Your WAF is set to observe mode — switching to throttle would free up workers immediately.
6 tools, one question away
Security Overview
Server counts, events today, active bans — your dashboard at a glance.
"How are my servers doing?"
Server Status
CPU, memory, disk, WAF config, and active ban count per server.
"Is smartsalus online?"
Search Events
Filter by type (brute_force, sql_injection, bot_detected...), severity, or server.
"Show me critical events from today"
Active Bans
All blocked IPs with reason, escalation count, and expiry time.
"Which IPs are permanently banned?"
Security Briefing
Full report: incidents, trends, WAF scoring, CVEs, and recommendations.
"What happened while I was sleeping?"
Server Details
Deep dive: recent events, active bans, firewall rules, resource metrics.
"What is attacking web-prod-01?"
Setup in 2 minutes
Create an API token, add one JSON block, done.
Create an API token
Go to Settings → API Tokens and click "Create Token".
Add to your AI client
// claude_desktop_config.json
{
"mcpServers": {
"defensia": {
"command": "npx",
"args": ["-y", "@defensia/mcp-server"],
"env": {
"DEFENSIA_API_TOKEN": "your-token-here"
}
}
}
}Ask anything
Open your AI assistant and ask: "Are my servers under attack?" — the MCP server handles the rest.
Works with any MCP client
Any AI assistant that supports the Model Context Protocol.
Claude Desktop
Desktop app
Claude Code
CLI
Cursor
IDE
Windsurf
IDE
Read-only by design
No write access
The MCP server can only read data from your dashboard. It cannot modify firewall rules, ban IPs, or change any server configuration.
Token stays local
Your API token lives in your local config file. It is only sent to the Defensia API — never to the AI model or any third party.
Zero telemetry
The MCP server collects no analytics, sends no telemetry, and phones home to nobody. It is fully open source.
Open source
Every line of code is on GitHub under MIT license. Audit it before installing.
How it works
┌─────────────────────────────────────────────────────┐
│ Your AI assistant │
│ (Claude Desktop, Claude Code, Cursor, Windsurf) │
└──────────────────────┬──────────────────────────────┘
│ MCP protocol (stdio)
▼
┌─────────────────┐
│ Defensia MCP │ ← npm package
│ Server (Node) │ runs locally
└────────┬────────┘
│ HTTPS + Bearer token
▼
┌─────────────────┐
│ Defensia API │ ← your dashboard
│ (REST + Auth) │ defensia.cloud
└────────┬────────┘
┌───────────┼───────────┐
▼ ▼ ▼
┌─────────┐ ┌─────────┐ ┌─────────┐
│ Agent 1 │ │ Agent 2 │ │ Agent N │ ← your servers
│ (VPS) │ │(Docker) │ │ (K8s) │
└─────────┘ └─────────┘ └─────────┘Start asking your servers
Create a free account, connect a server, and ask your first security question in under 3 minutes.