IDA Moles is a reverse analysis interface for IDA Pro 9.1. It controls decompilation, debugging, and other operations via standardized calls, returning POST-formatted results. It supports custom MCP server development to enhance reverse analysis efficiency and flexibility.
IDA Moles MCP exhibits pervasive quality gaps across naming, descriptions, and schemas. All 35 tools lack input parameter descriptions, a critical violation of pattern:tool-description. Descriptions are present but generic and often in Chinese, limiting LLM utility. No output schemas are documented. Parameter types are declared (Union[int, str]) but lack validation constraints, ranges, or format guidance. No error handling guidance, recovery paths, or actionable error messages. Tools are narrowly scoped (good), but composition is weak, many tools return JSON strings requiring LLM parsing rather than structured objects. No pagination, no idempotency hints, no permission gates.
将内存地址映射到反编译代码的行号
对已检查的函数进行反编译(特定场景下使用)
通过地址对函数进行反编译
通过函数名对函数进行反编译
将反编译代码的行号映射到内存地址
对指定地址范围内的指令进行反汇编
从指定地址开始反汇编指定行数的指令
All 35 tools lack parameter descriptions. Every parameter (func_start_addr, address, line_count, etc.) has a type but no description explaining its purpose, format, or constraints. This violates pattern:tool-description and forces LLMs to guess parameter semantics.
No output schemas documented. Tools return JSON strings (e.g., 'JSON字符串') but the structure is never specified. LLMs cannot plan downstream calls or extract fields without knowing the response schema.
Descriptions are generic and in Chinese. Many descriptions are under 50 characters and lack actionable context (e.g., 'get_basic_info' → '获取基础信息'). Non-English descriptions limit LLM utility in multilingual deployments.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 47 | 2026-07-28+ | v2 |
对指定地址的函数进行反汇编
通过关键字搜索函数(模糊匹配)
获取镜像中所有函数的列表信息 返回值:包含所有函数基本信息的JSON字符串
获取镜像中所有段信息(代码段、数据段、地址范围等) 返回值:包含所有段信息的JSON字符串
获取基础信息(如程序基本属性等) 返回值:包含基础信息的JSON字符串
获取已定义的数据结构信息 返回值:包含数据结构信息的JSON字符串
获取程序的所有入口点信息 返回值:包含入口点信息的JSON字符串
获取指向指定地址的第一个代码交叉引用
通过地址查找函数信息
通过函数名查找函数信息
获取镜像中的函数总数 返回值:包含函数计数的JSON字符串
获取指定地址函数的详细信息(参数、返回值、交叉引用等)
获取镜像文件的详细信息(架构、格式、入口点等) 返回值:包含镜像信息的JSON字符串
获取镜像中的导入函数列表(从外部库导入的函数) 返回值:包含导入函数信息的JSON字符串
读取指定地址的1字节数据
读取指定地址开始的连续字节数据
读取指定地址的4字节数据(双字)
读取指定地址的8字节数据(四字)
读取指定地址的2字节数据(字)
获取指定地址的微代码信息
获取镜像中的段总数 返回值:包含段计数的JSON字符串
获取指定地址所属的段信息
获取当前选中的反编译代码 返回值:包含选中反编译代码的JSON字符串
获取当前选中的汇编代码 返回值:包含选中汇编代码的JSON字符串
获取当前选中区域的十六进制数据 返回值:包含十六进制数据的JSON字符串
获取程序中的字符串信息 返回值:包含字符串信息的JSON字符串
通过类型名获取类型信息
在指定地址范围内搜索数据
No error handling or recovery guidance. Tools return JSON strings but no documentation of error cases, failure modes, or what the LLM should do if a call fails (retry, ask user, abort).
No pagination or result limits. Tools like get_all_functions, get_all_segments, get_string_info return unbounded lists. Large results will exhaust context windows and degrade LLM reasoning.