内容
# Burp MCP Agents
Practical setup guides and helpers to connect Burp Suite MCP Server to multiple AI backends (Codex, Gemini, Ollama, LM Studio).
This repo focuses on wiring, safety-first workflows, and reusable prompts to analyze real Burp traffic.
No fuzzing.
No blind scanning.
Only real traffic + reasoning.
---
## What this is
Burp MCP Agents is a collection of:
• Backend setup guides (Codex, Gemini, Ollama, LM Studio)
• Caddy proxy configuration for MCP SSE
• Prompt templates for passive analysis and reporting
---
## Architecture
```yaml
Burp Suite MCP Server
▲
│
MCP Bridge
│
┌──────────────────────────────────────────────┐
│ │ │ │
Codex CLI Ollama Agent Gemini CLI LM Studio Agent
```
---
## Supported Backends
| Backend | Mode | Privacy | Difficulty |
|--------|-----|--------|------------|
| Codex CLI | Cloud | Medium | Easy |
| Ollama | Local | Full | Advanced |
| Gemini CLI | Cloud | Medium | Easy |
| LM Studio | Local | Full | Advanced |
---
## Quick start
All setups require:
1. Burp MCP Server plguin enabled
2. Caddy reverse proxy (see `common/caddy_setup.md`)
3. One backend of your choice
### Install the Burp MCP Server extension
1) Download the Burp MCP Server extension (MCP server jar) from:
https://portswigger.net/bappstore/9952290f04ed4f628e624d0aa9dccebc
2) In Burp Suite: `Extender` → `Extensions` → `Add` → select the jar or from BApp Store.
3) Start the extension and confirm it listens on `127.0.0.1:9876`.
---
## Codex CLI
See: `codex/README.md`
**Example models**
| Model | Use |
|------|----|
| gpt-5.2-codex | General use |
| gpt-5.1 | Faster |
| gpt-5-mini | Low resource |
---
## Ollama (fully local)
See: `ollama/README.md`
**Example models**
| Model | VRAM | Notes |
|------|-----|------|
| llama3.1:8b-instruct | 8GB+ | Small, fast |
| qwen2.5:14b-instruct | 16GB | Mid size |
| llama3.1:70b-instruct | 48GB+ | Large, high VRAM |
---
## Gemini CLI
See: `gemini-cli/README.md`
**Example models**
| Model | Notes |
|------|-----|
| gemini-2.0-flash | Fast |
| gemini-2.0-pro | Deeper reasoning |
---
## LM Studio (local OpenAI-compatible)
See: `lmstudio/README.md`
**Example models**
| Model | Notes |
|------|-----|
| llama-3.1-8b-instruct | Small, fast |
| qwen2.5-14b-instruct | Mid size |
| llama-3.1-70b-instruct | Large, high VRAM |
---
## Prompts
The real power lives in `prompts/`:
| Prompt | Purpose |
|------|--------|
| passive_hunter.md | Broad passive vuln surfacing |
| idor_hunter.md | IDOR/BOLA discovery |
| auth_flow_mapper.md | Auth vs unauth access mapping |
| ssrf_redirect_hunter.md | SSRF/open redirect candidates |
| logic_flaw_hunter.md | Multi-step logic issues |
| session_scope_hunter.md | Token scope/audience misuse |
| rate_limit_abuse_hunter.md | Rate-limit and abuse gaps |
| report_writer.md | Evidence-based reporting |
See `prompts/README.md` for usage guidance.
---
## Optional launchers
You can use the backend launchers to auto-start Caddy and shut it down when the
backend exits.
Source them directly:
```bash
source /path/to/burp-mcp-agents/codex/burpcodex.sh
source /path/to/burp-mcp-agents/gemini-cli/burpgemini.sh
source /path/to/burp-mcp-agents/ollama/burpollama.sh
source /path/to/burp-mcp-agents/lmstudio/burplmstudio.sh
```
Then run:
```bash
burpcodex
burpgemini
burpollama deepseek-r1:14b
burplmstudio llama-3.1-8b-instruct
```
To make these available in every shell, add the `source` lines to your
`~/.zshrc`.
---
## What this enables
You are not running a scanner.
You are reviewing real traffic with assisted reasoning.
