re-speakeasy
MCP server for Speakeasy — Windows API emulation for binary analysis. It runs a .exe in a Wine-like sandbox and returns a structured API trace.
README
re-speakeasy
MCP server for Speakeasy (Mandiant, Apache-2.0) — Windows API emulation for binary analysis. The "run a .exe in a Wine-like sandbox and tell me what it did" tool.
Why
The other RE-AI MCP servers analyze binaries statically (no execution): re-lief reads headers, re-rizin disassembles, re-triton symbolically executes individual functions. None of them can answer "what APIs does this .exe call, with what arguments, in what order, when I run it?"
re-speakeasy fills that gap. Speakeasy is a Windows API emulator — it loads the .exe / .dll in-process and serves the same Win32 surface that Windows would, but in pure Python. The output is a per-API trace that complements the static analysis: "the static strings say this binary calls CreateFileW" + "the dynamic trace confirms it calls CreateFileW("C:\\Users\\...", GENERIC_READ, ...)".
Architecture
The Python MCP server is a thin wrapper around a speakeasy-cli Python helper installed by install.sh:
Claude Code (MCP stdio)
│
▼
re-speakeasy server (Python, this directory)
│ subprocess.run(...)
▼
speakeasy-cli (small Python script, wraps the Speakeasy API)
│
└─ speakeasy-emulator (pip-installed, the actual emulator)
The subprocess boundary is intentional: Speakeasy is a heavy Python package with rich in-process APIs. The subprocess wrapper keeps the MCP server's memory footprint small and lets Claude Code load the plugin in degraded mode if Speakeasy isn't installed.
Tools
| Tool | What it does |
|---|---|
check_speakeasy |
Health check — return Speakeasy version + API count |
emulate_binary |
Run a .exe / .dll under Speakeasy, return per-API trace |
list_emulated_apis |
Return the count + sample of the Win32 API catalog Speakeasy emulates |
Install
./install.sh installs speakeasy-emulator from PyPI.
To install standalone:
pip install speakeasy-emulator
Requirements
- Python 3.11+
speakeasy-emulator(Apache-2.0, on PyPI)- No system dependencies
Degraded mode
If speakeasy-cli is not installed, every tool returns {"status": "WARN", "error": "speakeasy-cli not installed; run install.sh", ...}. The Python MCP server itself always loads so Claude Code can surface the install hint.
Pairing with re-winedbg
re-winedbg runs a Windows .exe under Wine (the full Windows compatibility layer, including x86_64 emulation) and exposes a gdbserver for interactive debugging. re-speakeasy runs the .exe under Speakeasy (the pure-Python emulator, no real CPU) and returns a structured API trace.
- Use
re-speakeasy.emulate_binaryfor "what did this binary do, end-to-end, with a structured trace?" — fast, no x86 emulation, can be retried safely. - Use
re-winedbg.start_winedbg_gdbserverfor "I want to step through this binary interactively, with breakpoints" — slower (full Wine + gdbserver), but gives the analyst control.
For the encrypted-VM bytecode family: re-speakeasy.emulate_binary is the right first call (let the encrypted stub decrypt, watch the dispatcher fire, see which handlers execute); re-winedbg is the right follow-up when the analyst wants to break at a specific handler entry.
Pairing with re-leak-scan
The Speakeasy trace includes network calls (WinHttpOpen, WSAConnect, InternetOpenUrl, ...). Cross-reference against re-leak-scan.find_secrets to confirm whether the dynamic calls match the static string-table leaks. The Sentry DSN or Logstash URL in the strings is a credential; the same URL appearing in the Speakeasy trace is the actual call site.
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。