Kratos-MCP
Enables autonomous analysis of PE (Portable Executable) files for detecting injected sections and anomalies, with self-correcting JSON-RPC tools for binary forensics.
README
Kratos-MCP: Autonomous PE-Injected Section Analyzer
Terminal Layout Integrity & Section Mutation Detection Engine
Designed & Engineered for the SANS FIND EVIL Hackathon
📸 Dashboard Telemetry Visualizer
Below is the high-fidelity military-grade cyber tactical TUI dashboard for Kratos-MCP. This interactive dashboard runs entirely inside your terminal, tracking live PE structure telemetry, alerting on mutated segments, and displaying agent self-correction steps.

⚡ Quick Start & Installation Guide
Kratos-MCP is a zero-dependency, single-script application written in pure Python. It requires no external package installations or third-party web frameworks. It runs directly on the standard SANS SIFT Workstation.
1. Running the Interactive Terminal TUI
To launch the interactive terminal dashboard:
python3 kratos_mcp.py
- Keyboard Interaction:
- Press
[A]to execute the simulated autonomous agent forensic audit (runs log streaming animations). - Press
[1-4]keys to select and inspect PE sections in real-time. - Press
[E]to export the forensic state snapshot JSON. - Press
[Q]to exit the interface.
- Press
2. Connecting to the JSON-RPC MCP Server
Kratos-MCP implements a standard Model Context Protocol (MCP) server over stdio. It interfaces with autonomous agents (e.g. Claude Code or Cursor) when stdin/stdout are piped.
MCP JSON Configuration Example:
To hook Kratos-MCP into your Claude Code workspace agent, add this configuration block to your client's mcp.json settings:
{
"mcpServers": {
"kratos-mcp": {
"command": "python3",
"args": ["/absolute/path/to/kratos_mcp.py"]
}
}
}
Manual Stdio JSON-RPC Testing:
To test the JSON-RPC interface manually:
- Run the script:
python3 kratos_mcp.py(with stdin piped). - Paste the following JSON-RPC commands on stdin:
- List Tools:
{"jsonrpc": "2.0", "id": 1, "method": "tools/list"} - Parse Headers:
{"jsonrpc": "2.0", "id": 2, "method": "tools/call", "params": {"name": "parse_pe_sections"}} - Trigger Self-Correction Error Check:
{"jsonrpc": "2.0", "id": 3, "method": "tools/call", "params": {"name": "extract_section_bytes", "arguments": {"section_name": "data"}}} - Re-execute Corrected Query:
{"jsonrpc": "2.0", "id": 4, "method": "tools/call", "params": {"name": "extract_section_bytes", "arguments": {"section_name": ".data"}}}
- List Tools:
🌟 15 Unique Enterprise Features
Kratos-MCP is built from the ground up to support deep binary forensics:
- DKOM ActiveProcessLinks Double-Traversal: Verifies process list structures by traversing forward and backward pointers.
- SSDT (System Service Descriptor Table) Hook Detection: Scans core Windows kernel system calls (e.g.
NtQuerySystemInformation) to identify rootkit redirects. - IDT (Interrupt Descriptor Table) Mutation Alert: Audits core interrupt vectors to catch persistence mechanisms.
- Kernel Mode Driver Signature Scanner: Flags unsigned or forged certificate drivers (
.sysfiles) loaded in Ring 0. - Memory Page Permissions Auditor: Audits page table protections (e.g. tracking executable allocations in writable space).
- Token Elevation Tracker: Flags modifications to system process security tokens (e.g. unauthorized elevation to
NT AUTHORITY\SYSTEM). - Automated SANS DFIR Timeline Aggregator: Automatically maps parsed timestamp artifacts into standard investigation audits.
- Asynchronous Multi-Agent Telemetry Stream: Formats output telemetry for consumption by autonomous multi-agent systems.
- Hexadecimal Memory Address Constraint Evaluator: Dynamic validator checking address spaces and alignments.
- Rogue Thread Call-Stack Inspector: Extracts thread stacks to identify hollowing or DLL injection vectors.
- Cryptographic Chain of Custody Validation: Immutably seals session states using SHA-256 tokens to ensure evidence tracking compliance.
- Self-Correction Logic Routing Framework: Returns explicit schema hints on formatting failures so agents can self-correct.
- VAD (Virtual Address Descriptor) Tree Cross-Referencer: Cross-matches virtual address mappings with active process lists.
- Network Socket Port Socket Correlator: Maps open TCP/UDP sockets directly to active processes.
- One-Click Forensics State Snapshot Exporter: Serializes the parsed PE structures into a signed JSON snapshot.
⚙️ Technical Blueprint & SANS Compliance
Kratos-MCP satisfies SANS incident response criteria through three core mechanics:
1. Self-Correction Logic Routing
To verify the agent's reasoning, the extract_section_bytes tool strictly validates parameters. If the agent requests bytes from a section using raw names without dot notation (e.g., data instead of .data), the server throws an explicit schema error. This triggers a self-correction loop where the agent parses the error hint, reformats the string to .data, and retries the tool call successfully.
2. Absolute Accuracy Validation
Every virtual offset, file header address, and section entropy value is mapped directly to actual low-level binary headers. All outputs are stamped with a cryptographically generated session_token to guarantee evidence tracking chain-of-custody.
3. Analytical Reasoning
Instead of returning long, unstructured log sequences, the final output compiles directly into a SANS DFIR executive case file layout, translating low-level anomalies into an actionable threat intelligence summary automatically.
📜 MIT License
This project is licensed under the MIT License - see the LICENSE file for details.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。