mcp-hayabusa
Enables scanning Windows EVTX event log files with Hayabusa, returning structured detection results through an MCP tool.
README
mcp-hayabusa
An MCP server that wraps Hayabusa for EVTX (Windows event log) analysis, exposing scan_evtx and get_hayabusa_rules tools to Claude.
Setup
-
Install dependencies:
pip install -r requirements.txt -
Download the Hayabusa binary and Sigma rules into
./hayabusa/:python download_hayabusa.pyThis fetches the latest release for your OS/architecture and extracts it to
./hayabusa/(binary at./hayabusa/hayabusa.exeon Windows or./hayabusa/hayabusaelsewhere, plusrules/andrules/config/). -
Keep the ruleset current (recommended before each scanning session):
./hayabusa/hayabusa.exe update-rules # Windows ./hayabusa/hayabusa update-rules # Linux/macOS
Running the server
python server.py
This starts the MCP server over stdio. Point an MCP client at it, e.g. in Claude Desktop's config:
{
"mcpServers": {
"hayabusa": {
"command": "python",
"args": ["C:\\path\\to\\mcp-hayabusa\\server.py"]
}
}
}
Tool: scan_evtx
Scans an EVTX file (or a directory of EVTX files) with Hayabusa and returns detections as structured JSON.
Parameters:
evtx_path(string, required) — path to a.evtxfile or a directory containing themmin_severity(string, optional, default"informational") — minimum severity to include:informational,low,medium,high,criticalrule_filter(string, optional) — case-insensitive substring match against each detection's rule title (e.g."lateral"or"mimikatz"); only matching detections are returned. Hayabusa has no native rule-title filter, so this is applied after the scan.output_format(string, optional, default"summary") —"summary"returns condensed detections (Timestamp,RuleTitle,Level,Computer,Channel,EventID,RecordID);"full"includes the completeDetails/ExtraFieldInfopayload for each detectionmax_results(integer, optional) — caps the number of detections returned, applied afterrule_filter
Returns:
{
"evtx_path": "...",
"min_severity": "...",
"rule_filter": "...",
"output_format": "...",
"total_count": 68,
"count": 42,
"truncated": false,
"detections": [ { "Timestamp": "...", "RuleTitle": "...", "Level": "...", "...": "..." } ]
}
total_count is the number of matching detections before max_results is applied; count is the number actually returned; truncated is true if max_results cut off results.
On failure (missing file, missing Hayabusa binary, invalid min_severity/output_format/max_results, scan timeout, or a Hayabusa scan error), it returns {"error": "..."} instead of raising.
Tool: get_hayabusa_rules
Lists available Hayabusa/Sigma detection rules from ./hayabusa/rules/, optionally filtered by keyword. Useful for seeing what detections exist before scanning, or for finding a good rule_filter value for scan_evtx. Hayabusa has no built-in rule-listing command, so this reads and parses the rule YAML files directly.
Parameters:
keyword(string, optional) — case-insensitive substring matched against each rule's title, description, tags, and id. If omitted, all rules are listed (subject tomax_results)max_results(integer, optional, default100) — caps the number of rules returned
Returns:
{
"keyword": "...",
"total_count": 66,
"count": 66,
"truncated": false,
"rules": [
{
"id": "...",
"title": "...",
"level": "...",
"status": "...",
"description": "...",
"logsource": { "product": "windows", "service": "..." },
"tags": ["attack.lateral-movement", "..."],
"file": "hayabusa\\builtin\\System\\Sys_7045_Med_LateralMovement-PSEXEC.yml"
}
]
}
total_count is the number of matching rules before max_results is applied; count is the number actually returned; truncated is true if max_results cut off results.
A keyword search checks a raw-text prefilter before parsing each rule's YAML, so it typically runs in ~1-2 seconds. Listing all ~5,000 rules with no keyword takes several seconds longer since every rule file must be parsed.
On failure (missing rules directory or invalid max_results), it returns {"error": "..."} instead of raising.
Requirements
- Python with the
mcpandpyyamllibraries (seerequirements.txt) - The Hayabusa CLI, installed via
download_hayabusa.py
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。
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。
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。
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。
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。
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。
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。