The Crash MCP Server is a unified platform for analyzing Linux system crash dumps by integrating the crash utility and drgn programmable debugger through an MCP interface.
Core Capabilities:
Crash dump analysis: Analyze system crash dumps (vmcore files) using crash utility commands and drgn Python debugging
Session management: Start, manage, and terminate analysis sessions with automatic deduplication and reuse for the same vmcore
Remote analysis: Connect to remote hosts via SSH to analyze crash dumps without downloading files locally
Command execution: Run crash commands (
sys,bt,log,ps,files,vm,kmem) and drgn Python code for advanced debuggingFile discovery: Recursively scan directories to locate crash dump files
Output management: Persist command outputs to disk with configurable truncation, pagination support, and regex search capabilities
Multi-architecture support: Automatically detect vmcore architecture and compile/use appropriate crash versions (x86_64, ARM64) with support for different compression formats like LZO
Command caching: Cache outputs for improved performance on repeated queries
Dual transport modes: Operate in stdio (default) or SSE/HTTP modes for different deployment scenarios
Configuration flexibility: Customize behavior through environment variables for caching, logging, and output truncation
Provides tools for analyzing Linux system crash dumps through the crash utility, enabling automatic discovery of vmcore files, kernel matching, and execution of crash analysis commands like sys, bt, and ps.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Crash MCP Serveranalyze the latest crash dump and show the backtrace"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Crash MCP Server
基于 MCP (Model Context Protocol) 的服务器，用于 系统崩溃转储 (Crash Dump) 分析。
集成 Linux crash 实用程序和 drgn 可编程调试器，提供统一分析接口。
功能特性
统一会话: 同时支持
crash和drgn引擎会话去重: 同一 vmcore 自动复用已有会话
命令持久化: 输出自动落盘，支持分页和搜索
远程分析: 通过 SSH 连接远程主机，无需下载 vmcore
多传输模式: Stdio（默认） / SSE（HTTP）
智能架构识别: 自动检测 vmcore 架构并选择对应的 crash 版本
自动编译: 内置 crash 工具编译器，支持多架构和压缩格式
安装
前置要求
Python 3.10+
crash工具（可通过内置编译器安装）python3-dev(编译 PyKdump 需要)drgn工具（通过 pip 自动安装）
快速安装
chmod +x install.sh && ./install.sh编译 Crash 工具
# 激活虚拟环境
source venv/bin/activate
# 查看依赖安装说明
compile-crash --deps
# 编译 x86_64 版本
compile-crash
# 编译 ARM64 版本（在 x86_64 上分析 ARM64 vmcore）
compile-crash --arch ARM64
# 编译带 PyKdump 支持的版本
compile-crash --pykdump-from-source使用
启动服务器
# Stdio 模式
crash-mcp
# SSE 模式
crash-mcp --transport sse --port 8000MCP 工具
工具 | 说明 |
| 打开 vmcore 崩溃转储文件进行分析 |
| 执行 crash 命令，支持 PyKdump 扩展 |
| 执行 drgn Python 代码 |
| 关闭当前分析会话 |
| 分页获取长命令输出 |
| 正则搜索命令输出 |
| 运行预定义分析脚本 (需配置 |
| 列出可用分析脚本 (需配置 |
| 获取崩溃诊断报告 (需配置 |
配置
环境变量 | 默认值 | 说明 |
|
| 是否自动加载扩展 |
|
| 输出截断行数 |
|
| 会话工作目录 |
|
| 启用命令缓存 |
|
| 日志级别 |
| (空) | 自动化诊断脚本命令模板 (如 |
| (空) | 外部 drgn 脚本搜索路径 (冒号分隔) |
客户端配置
Claude Desktop (claude_desktop_config.json):
{
"mcpServers": {
"crash-analysis": {
"command": "/path/to/crash-mcp/venv/bin/crash-mcp"
}
}
}示例
# 本地分析
open_vmcore_session("/var/crash/vmcore", "/usr/lib/debug/vmlinux")
run_crash_command("bt")
run_crash_command("sys")
run_drgn_command("prog['init_task'].comm")
# 分页获取长输出
run_crash_command("ps") # 返回 command_id
get_command_output("crash:ps", offset=20, limit=50)
# 搜索输出
search_command_output("crash:bt", "schedule")
# 远程分析
open_vmcore_session("/var/crash/vmcore", "/usr/lib/debug/vmlinux",
ssh_host="server-01", ssh_user="root")常见问题
1. no lzo compression support 错误
vmcore 文件使用 LZO 压缩，但 crash 工具编译时未启用 LZO 支持。
sudo apt-get install liblzo2-dev
compile-crash --clean2. 缺少 GMP/MPFR 库
sudo apt-get install libgmp-dev libmpfr-dev3. 查看所有编译依赖
compile-crash --deps许可证
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.