{
  "added_at": "2026-11-06",
  "additional_evidence_urls": [
    "https://github.com/MeroZemory/ida-multi-mcp/blob/8ff883109fc3ce614d1e59718a53b02a0634c542/pyproject.toml",
    "https://github.com/MeroZemory/ida-multi-mcp/blob/8ff883109fc3ce614d1e59718a53b02a0634c542/docs/installation.md"
  ],
  "category": "agent-mcp",
  "description_en": "MCP server that lets an agent analyze several IDA Pro databases at once through one endpoint. Each database with its plugin registers itself when opened and headless idalib sessions can be added, and a router exposes up to 76 tools for triage, decompilation and disassembly, cross-references and call graphs, search, types, renaming, comments and patching, plus analysis-wait gating so agents can wait for IDA's auto-analysis to settle; its IDA tool implementations come from the ida-pro-mcp project. Function-similarity search compares functions within or across binaries using name-independent instruction MinHash, API, string and constant anchors and control-flow shape, with optional local neural recall, and nothing is uploaded.",
  "description_zh": "让智能体通过同一个端点同时分析多个 IDA Pro 数据库的 MCP 服务器。装有其插件的数据库在打开时会自动注册，也可加入无界面的 idalib 会话；路由器最多提供 76 个工具，涵盖初步分析、反编译与反汇编、交叉引用与调用图、搜索、类型、重命名、注释和补丁，并提供等待自动分析完成的机制，让智能体在 IDA 分析稳定后再下结论；其 IDA 工具实现来自 ida-pro-mcp 项目。函数相似度搜索可在单个二进制内或跨二进制比较函数，依据不依赖函数名的指令 MinHash、API/字符串/常量锚点和控制流结构，并可选用本地神经网络召回，不会上传任何数据。",
  "entry_url": "https://awesomerea.org/resources/ida-multi-mcp/",
  "entry_url_zh": "https://awesomerea.org/zh/resources/ida-multi-mcp/",
  "evidence_url": "https://github.com/MeroZemory/ida-multi-mcp/blob/8ff883109fc3ce614d1e59718a53b02a0634c542/README.md",
  "group": "agents",
  "id": "ida-multi-mcp",
  "json_url": "https://awesomerea.org/resources/ida-multi-mcp/record.json",
  "kind": "integration",
  "markdown_url": "https://awesomerea.org/resources/ida-multi-mcp/index.md",
  "markdown_url_zh": "https://awesomerea.org/zh/resources/ida-multi-mcp/index.md",
  "name": "ida-multi-mcp",
  "name_zh": "ida-multi-mcp 多实例 IDA Pro MCP 服务器",
  "record_type": "resource",
  "relationship": "independent-agent-integration",
  "requirement_facets": [
    "prerequisites",
    "limits",
    "license"
  ],
  "requirements": [
    "Needs Python 3.11 or later and IDA Pro 8.5 or later (9.1 or later recommended); IDA 9.0 SP0 (build 240925) is not supported. The package must be installed for both the terminal's Python and IDA's Python, and headless idalib sessions need IDA Pro, because IDA Home and IDA Free do not ship idalib. The repository has no releases; pyproject.toml declares version 0.1.0, and the reviewed main commit dates from 2026-09-19.",
    "The README lists limits: the router binds to localhost only and speaks only stdio, calls to the same IDA instance still run one at a time, cancellation is not forwarded to IDA, and resources need manual routing. Renames, retypes and comments stay in memory until idb_save is called. The optional neural recall downloads its model on demand.",
    "MIT per the README and pyproject.toml; the repository has no LICENSE file."
  ],
  "requirements_zh": [
    "需要 Python 3.11 或更高版本，以及 IDA Pro 8.5 或更高版本（推荐 9.1 及以上）；不支持 IDA 9.0 SP0（构建号 240925）。软件包须同时安装到终端所用的 Python 和 IDA 自带的 Python 中；无界面的 idalib 会话需要 IDA Pro，因为 IDA Home 和 IDA Free 不含 idalib。仓库没有发布版本，pyproject.toml 标注的版本为 0.1.0，所审阅的 main 分支提交日期为 2026-09-19。",
    "README 列出的局限：路由器只绑定本机回环地址、只支持 stdio 传输，对同一个 IDA 实例的调用仍须排队执行，取消请求不会转发给 IDA，资源（resources）需手动路由。重命名、修改类型和注释在调用 idb_save 之前只保存在内存中。可选的神经网络召回功能会按需下载模型。",
    "依 README 和 pyproject.toml 采用 MIT 许可；仓库中没有 LICENSE 文件。"
  ],
  "targets": [
    "binary"
  ],
  "url": "https://github.com/MeroZemory/ida-multi-mcp",
  "verification": "source-reviewed",
  "verified_at": "2026-11-05"
}