drogonclaw - An autonomous, AI-powered offensive security frame

项目说明

DrogonClaw

<di="center">
<img src="assets/logo.png" alt="DrogonClaw Logo" width="300" />
</div>

<di="center">

![CI]()
![Go]()

</div>

> Autonomous AI-Powered Penetration Testing Platform

DrogonClaw is an autonomous AI agent that plans, executes, and adapts penetration tests against target infrastructure. It operates as a ReAct-based intelligence core with persistent memory, parallel subagents, learned attack patterns, and sandboxed tool execution.

Quick Start | Capabilities | Architecture | Tools | Commands | De | Setup Guide

---

Warning

> [!WARNING]
> Linux Only — Authorized Testing Only
> DrogonClaw is designed for Linux-based operating systems (Kali, Ubuntu, Debian). It relies on Linux-specific networking APIs, filesystem permissions, and process management. Never use it against systems without explicit written authorization.

---

Capabilities

Core Engine

Intelligence & Exploitation

New Capabilities (Scrapling + Hermes Inspired)

Safety & Control

Interfaces

---

Quick Start

Requirements

Install (From Source)

Install (From npm)

Configure

Run the interacti wizard:

The wizard configures:
1. Authorization — Scope/compliance acknowledgement
2. Neural Provider — OpenRouter, NVIDIA NIM, OpenAI, Google Gemini, or Ollama
3. Credentials — Provider API key
4. Remote C2 (optional) —
5. Secondary API Keys (optional) — GitHub, Shodan, , Bra, Hunter.io, Exa

Configuration stored in ~/.drogonclaw/config.json (owner read/write only).

Start

Daemon Mode (Headless)

---

Supported Pro

| Provider | Config Key | Notes |
| --- | --- | --- |
| OpenRouter | openrouter | Multi-model gateway; li list with curated fallback |
| NVIDIA NIM | nvidia | High-performance inference (Nemotron, Qwen, DeepSeek, Llama) |
| OpenAI | openai | Direct API (gpt-4o, gpt-4o-mini) |
| Google Gemini | gemini | Enterprise reasoning (gemini-2.5-pro, gemini-2.5-flash) |
| Ollama | ollama | Offline runtime; set OLLAMA_BASE_URL |
| 9Router.ai | 9router | Intelligent routing with cost optimization (docs/ROUTING.md) |

---

Architecture

For detailed architecture documentation, see ARCHITECTURE.md.

---

Commands

Keyboard Shortcuts

| Shortcut | Action |
| --- | --- |
| Ctrl+P | Command palette |
| Ctrl+A | Toggle autopilot |
| Ctrl+B | Toggle sidebar |
| Ctrl+S | Show status |
| Ctrl+D | Show cost |
| Ctrl+E | Open pager |
| Ctrl+Y | Copy output |
| Ctrl+T | Toggle tool detail panel |
| Ctrl+C | Abort execution |
| Ctrl+X then b/n/l/m/t/e/x/q | Leader key commands |

Slash Commands

All commands are listed the same way /help renders them — one registry, zero drift.

Operations — Acti
| Command | Description |
| --- | --- |
| /workflow <name> (/mode alias) | Select attack workflow (recon/exploit/ctf/web/api/mail) |
| /analyze <target> | Classify target and suggest attack path |
| /profile <target> | Build passi profile |
| /ctf <path> | Sol CTF challenge |
| /report | Generate penetration test report |
| /swarm <objective> | Dispatch parallel sub-agent swarm |
| /skills [query] | List/search a modules |

Controls — Settings
| Command | Description |
| --- | --- |
| /config [set KEY VALUE] | View or modify settings |
| /set <KEY> <VALUE> | Quickly set a config value (alias: /config set) |
| /router [auto\|local\|9router\|off\|status] | Configure intelligent routing |
| /pro | Provider health & routing status dashboard |
| /theme [name] | Switch color theme (dark/light/dracula/nord/gruvbox) |
| /auto [on\|off] | Toggle autopilot — auto-accept long-running tools (Ctrl+A) |
| /set E [high\|medium\|low] | Runtime: OPSEC alias — rate-limiting & e |
| /set ISOLATION [on\|off] | Runtime: SANDBOX alias — Docker isolation |

Session — Metrics & Lifecycle
| Command | Description |
| --- | --- |
| /help | Command reference |
| /status | Session metrics: runtime, tools run, findings, phase |
| /cost | Token usage breakdown with routing savings |
| /health | En diagnostics: Docker, dependencies, sandbox |
| /timeline | Execution log: when each tool ran, duration, results |
| /findings | Detection summary: vulns, creds, flags, info |
| /setup | Configuration wizard |
| /new | Clear session memory (confirms yes/no) |
| /resume | Resume interrupted execution from last checkpoint |
| /copy | Export transcript to clipboard & file |
| /clear | Clear terminal output |
| /exit /quit | Terminate session gracefully |

UI — Display
| Command | Description |
| --- | --- |
| /sidebar | Toggle sidebar panel (Ctrl+B) |
| /details | Toggle tool detail panel (Ctrl+T) |

---

De

Repository Structure

---

Documentation

| Document | Description |
| --- | --- |
| ARCHITECTURE.md | Detailed architecture with all subsystems |
| docs/TUI.md | TUI guide, shortcuts, themes, commands |
| docs/ROUTING.md | Intelligent routing (local + 9Router.ai) |
| docs/PRO.md | Provider setup and switching guide |
| docs/TOOLS.md | Complete tool reference catalog |
| docs/SESSION_MANAGEMENT.md | Sessions, throttle, cache, WAF detection |
| docs/SKILL_LEARNING.md | Learned attack patterns system |
| docs/SUBAGENTS.md | Parallel execution framework |
| docs/setup.md | Detailed setup guide |
| CONTRIBUTING.md | Contribution guidelines |
| SECURITY.md | reporting |
| CHANGELOG.md | history |

---

Security

DrogonClaw is built for authorized security testing only. Report via SECURITY.md. Ne against networks without explicit written consent.

Contributing

See CONTRIBUTING.md. Ensure make build, make lint, and make test pass before submitting PRs.

Disclaimer

DrogonClaw is designed for authorized security testing only. Always ensure you ha permission before testing any system. Unauthorized access to computer systems is illegal.

License

GNU AGPL v3

---

Star History

!Star History

项目信息

开发语言：Go

Star 数：12

技术要点

该项目属于安全软件方向的开源资源，代码结构清晰，适合安全研究人员学习参考。项目围绕An autonomous, AI-powered offensive security framework. DrogonClaw uses an unrestricted ReAct engine to dynamically chain reconnaissance, exploitation, and post-exploitation workflows within isolated, ephemeral sandboxes.展开，提供了可运行的实现，对理解相关安全机制、工具原理与攻防技术具有较高的研究价值。

从技术实现角度看，此类项目的核心在于对底层原理的深入理解与工程化落地。开发者在研究过程中可以结合官方文档与社区讨论，梳理项目的初始化流程、核心模块职责与数据流转，逐步掌握其中的技术细节与设计思路。

使用与学习建议

1. 下载源码压缩包，解压到本地环境。

2. 根据项目类型准备对应的运行环境与依赖。

3. 按照 README 中的步骤完成配置与部署。

4. 在隔离的测试环境中运行验证，避免影响生产系统。

免责声明

本资源来源于开源社区，仅供安全研究、技术学习与合法授权范围内的攻防演练使用，请勿用于任何非法或侵害他人权益的用途。使用本项目产生的一切后果由使用者自行承担。