Test WAF security interactively with an AI agent and WAFtester MCP

What it does

A conversational AI agent that connects to WAFtester via MCP (Model Context Protocol) for interactive Web Application Firewall security testing. Type natural language requests — the agent picks the right tools, runs the tests, and explains the results.

About WAFtester

WAFtester is an open-source CLI for testing Web Application Firewalls. It ships 27 MCP tools, 2,800+ attack payloads across 18 categories (SQLi, XSS, SSRF, SSTI, command injection, XXE, and more), detection signatures for 26 WAF vendors and 9 CDNs, and enterprise-grade assessment with F1/MCC scoring and letter grades (A+ through F).

• GitHub: github.com/waftester/waftester
• Docs: Installation | Examples | Commands

Who it's for

• Security engineers running ad-hoc WAF assessments
• Penetration testers who want AI-assisted reconnaissance and bypass discovery
• DevSecOps teams validating WAF coverage before and after deployments
• API security teams testing OpenAPI/Swagger specs against WAF rules

How it works

The workflow has four nodes:

1. Chat Trigger — Opens an n8n chat interface where you type requests in plain English
2. AI Agent — Receives your message, reasons about which tools to call, and orchestrates the testing workflow
3. OpenAI Chat Model — Provides the LLM reasoning layer (GPT-4o recommended; swappable for Anthropic, Ollama, etc.)
4. WAFtester MCP — Connects to the WAFtester server via SSE and exposes all 27 tools to the agent

The agent follows a standard WAF testing workflow:

1. detect_waf — Fingerprint the WAF vendor and CDN protecting the target
2. discover — Map the attack surface (endpoints, parameters, technologies) from robots.txt, sitemaps, JavaScript, and Wayback Machine
3. learn — Generate a prioritized test plan based on discovery results
4. scan — Fire 2,800+ attack payloads and measure detection vs. bypass rates
5. bypass — Systematic mutation matrix testing to find WAF evasion techniques
6. assess — Generate a formal security grade with F1, precision, MCC, and false positive rate

Long-running operations (scan, assess, bypass, discover, discover_bypasses, event_crawl, scan_spec) run asynchronously — the agent polls for results automatically.

Key capabilities

Example prompts

• "What WAF is protecting https://example.com?"
• "Scan https://example.com for SQLi and XSS"
• "Find WAF bypasses for https://example.com"
• "Run a full security assessment of https://example.com"
• "Validate my API spec at https://example.com/openapi.json"
• "Discover the attack surface of https://example.com"

How to set up

1. Start WAFtester MCP server: docker run -p 8080:8080 ghcr.io/waftester/waftester:latest mcp --http :8080
2. Add OpenAI credentials in n8n: Settings → Credentials → New → OpenAI API
3. Select the credential in the OpenAI Chat Model node
4. Activate the workflow and open the chat interface

Alternatively, use the included docker-compose.yml to run both n8n and WAFtester together with docker compose up -d.

Requirements

Links

• WAFtester website
• GitHub repository
• Installation guide
• Full examples
• Docker Hub