Burp Suite for AI Agent

📁 Structure

burpsuite-for-ai-agent/
├── src/
│   └── index.js          ← MCP server (HTTP bridge + 11 MCP tools)
├── plugin/
│   └── burpAI.py         ← Jython 2.7 Burp plugin (context menu, auto-forward, outbound polling)
├── package.json          ← Node.js manifest (dependencies: @modelcontextprotocol/sdk, zod)
├── .gitignore
└── README.md

Related MCP server: Burp Suite MCP Server

Architecture

Burp Suite (Linux / Windows)
  └─ plugin/burpAI.py (Jython plugin)
       │  POST /ingest (127.0.0.1:9999)
       │  GET /burp/outbound (poll every 2s)
       ▼
src/index.js  ← MCP server (HTTP bridge + 11 MCP tools)
       │
       ▼
AI Agent ←→ MCP tools

Linux: Burp + server + agent — all on one machine. Windows + WSL: Burp on Windows, server + agent in WSL over localhost.

Quick Install (one command)

bash install.sh

This runs npm install, prints Burp plugin instructions, and generates .mcp.json for AI agent auto-registration.

Quick Start

1. MCP server

node src/index.js

2. Load Burp plugin

1. Burp Suite → Extensions → Installed → Add

2. Extension type: Python (requires Jython 2.7 standalone JAR)

3. File: plugin/burpAI.py

3. Verify

In Burp's burpAI tab, click Check Status. Connected:

{ "ok": true, "requests": 0, "endpoints": 0 }

MCP Tools (16 tools)

Workflow

Capture → Analyze

Right-click a request in Burp (Proxy/Repeater) → burpAI: send request(s)

Agent:

• burp_requests — list captured requests

• burp_endpoints — enumerate endpoints

• burp_request_detail — full request/response

Queue scan → Execute

Right-click → burpAI: send + queue scan
Agent picks up via burp_tasks.

Finding → Burp Scanner

Agent calls burp_import_issue → Burp tab → Import Issues

Agent → Burp Repeater

Agent calls burp_send_to_burp with type send_to_repeater → Burp opens Repeater tab.

Agent-Driven Scanning

This is the core custom scanner flow: Burp captures → agent thinks → agent modifies → Burp executes.

Burp (Proxy → History)
  │ right-click → burpAI: send request(s)
  ▼
MCP server ←── captures request
  │
  ▼
Agent reads via burp_requests / burp_request_detail
  │
  │ Agent thinks: "endpoint /api/login takes username, try SQLi"
  │
  ▼
Agent calls burp_replay({
  request_id: "burp:burp-12345",
  set_body: '{"username":"admin\\' OR 1=1--","password":"x"}'
})
  │
  ▼
MCP server → Burp Repeater → Burp executes HTTP call
  │
  ▼ (auto-forward enabled)
Response captured back → Agent reads & analyzes
  │
  ▼
burp_import_issue → if vulnerable

Key tool: burp_replay takes a captured request ID, applies modifications (method, path, headers, body), and sends it to Burp Repeater. No need to manually craft raw bytes — the agent just specifies what to change.

Requirements:

• Enable Auto-send Repeater responses in Burp plugin settings

• This ensures the agent sees every Repeater response automatically

External Scanning (nuclei)

Optionally run nuclei scans directly from the bridge:

Nuclei must be installed on the same machine. Findings with High/Critical severity are auto-imported as Burp issues.

Auto-Forwarding

Options

node src/index.js --port 9999                 # Custom port (default: 9999)
node src/index.js --db /path/to/data.json     # Persistence file (default: burpai-data.json)
node src/index.js --port 9001 --db custom.json

Data persists across restarts when --db is set (or by default). The server auto-saves on every mutation and loads data on startup.

Burp plugin URL configurable from the burpAI settings tab in Burp.

Requirements

• Node.js 18+

• Burp Suite (Community or Professional)

• Jython 2.7 standalone JAR (configured in Burp Extensions → Environment)