{ "added_at": "2026-11-03", "additional_evidence_urls": [ "https://github.com/sjkim1127/Reversecore_MCP/blob/c1c576709e47d4947afec6e121111d44c036813c/pyproject.toml", "https://github.com/sjkim1127/Reversecore_MCP/blob/c1c576709e47d4947afec6e121111d44c036813c/SECURITY.md", "https://github.com/sjkim1127/Reversecore_MCP/blob/c1c576709e47d4947afec6e121111d44c036813c/AGENTS.md", "https://github.com/sjkim1127/Reversecore_MCP/blob/c1c576709e47d4947afec6e121111d44c036813c/ORIGINAL_REQUEST.md", "https://github.com/sjkim1127/Reversecore_MCP/blob/c1c576709e47d4947afec6e121111d44c036813c/Dockerfile", "https://github.com/sjkim1127/Reversecore_MCP/blob/c1c576709e47d4947afec6e121111d44c036813c/docker-compose.yml", "https://github.com/sjkim1127/Reversecore_MCP/blob/c1c576709e47d4947afec6e121111d44c036813c/server.json", "https://github.com/sjkim1127/Reversecore_MCP/blob/c1c576709e47d4947afec6e121111d44c036813c/.github/workflows/main.yml", "https://github.com/sjkim1127/Reversecore_MCP/tree/c1c576709e47d4947afec6e121111d44c036813c/tests/fixtures/benchmarks/targets", "https://github.com/sjkim1127/Reversecore_MCP/blob/c1c576709e47d4947afec6e121111d44c036813c/LICENSE", "https://github.com/sjkim1127/Reversecore_MCP/releases/tag/v3.0.5", "https://pypi.org/project/reversecore-mcp/3.0.5/" ], "category": "agent-mcp", "description_en": "MCP server that puts radare2 with r2ghidra, LIEF, Capstone, YARA, CAPA, Detect It Easy, binwalk, angr, Volatility3, Scapy, Sleuth Kit and other engines behind 177 typed tools for AI clients such as Claude and Cursor. The tools cover disassembly and decompilation, binary parsing, string, IOC and capability extraction, ESIL emulation and symbolic execution, YARA scanning, memory, PCAP and disk forensics, source-code pattern scanning and session reports with MITRE ATT&CK mapping, alongside 22 guided analysis prompts and per-binary MCP resources. File access is confined to a configured sample workspace, and the server runs over stdio or HTTP/SSE, typically from the project's Docker image.", "description_zh": "MCP 服务器，将 radare2 及 r2ghidra、LIEF、Capstone、YARA、CAPA、Detect It Easy、binwalk、angr、Volatility3、Scapy、Sleuth Kit 等引擎封装为 177 个带类型的工具，供 Claude、Cursor 等 AI 客户端调用。工具涵盖反汇编与反编译、二进制解析、字符串、IOC 与能力提取、ESIL 模拟与符号执行、YARA 扫描、内存、PCAP 与磁盘取证、源码模式扫描，以及带 MITRE ATT&CK 映射的会话报告，另有 22 种引导式分析提示词和按二进制划分的 MCP 资源。文件访问限定在配置的样本工作目录内，服务器通过 stdio 或 HTTP/SSE 运行，通常使用项目提供的 Docker 镜像。", "entry_url": "https://awesomerea.org/resources/reversecore-mcp/", "entry_url_zh": "https://awesomerea.org/zh/resources/reversecore-mcp/", "evidence_url": "https://github.com/sjkim1127/Reversecore_MCP/blob/c1c576709e47d4947afec6e121111d44c036813c/README.md", "group": "agents", "id": "reversecore-mcp", "json_url": "https://awesomerea.org/resources/reversecore-mcp/record.json", "kind": "integration", "markdown_url": "https://awesomerea.org/resources/reversecore-mcp/index.md", "markdown_url_zh": "https://awesomerea.org/zh/resources/reversecore-mcp/index.md", "name": "Reversecore MCP", "name_zh": "Reversecore MCP 逆向与安全分析 MCP 服务器", "record_type": "resource", "relationship": "independent-agent-integration", "requirements": [ "Latest release at review: v3.0.5, published 2026-09-28 on GitHub with a wheel, sdist, SBOMs and Sigstore bundles, on PyPI as reversecore-mcp and as a GHCR image. The reviewed main commit of 2026-10-17 is 147 commits ahead of that tag, while pyproject.toml still declares 3.0.5. The README's test and coverage figures (3,069 tests, 88%) are hard-coded badges and differ from the older counts in AGENTS.md.", "Python 3.10–3.12, with angr installed only on Python 3.12. Local installs need radare2 on the PATH, plus Volatility3, Scapy and Sleuth Kit for forensics; the README recommends the Docker image, which bundles the engines, with x86 and Apple Silicon compose profiles. It lists Linux or macOS hosts or any Docker environment, at least 4 cores, 8 GB RAM and 20 GB storage; Redis backs the task queue and result cache, and an SQLite memory database keeps notes across sessions.", "Binaries, memory images, PCAPs and disk images come from the analyst and are mounted into the workspace, where path validation confines tool access. Optional container sandboxing for dynamic-analysis tools is disabled by default, and HTTP mode accepts only loopback clients unless MCP_API_KEY is set. Some tools and prompts target exploit and PoC development.", "CI runs linting, type checks, Bandit, pip-audit and CodeQL, unit tests on Python 3.10–3.12 with an 85% coverage gate, security-boundary, integration and MCP client end-to-end suites, and a wheel build-and-install check; ClusterFuzzLite and nightly fuzzing run separately. Test fixtures include small synthetic vulnerable C targets with seeds, proof-of-concept inputs and ASan logs, and scratch/binaries holds only a hello.c source.", "MIT; the wrapped engines and the packages in the Docker image keep their own licenses. ORIGINAL_REQUEST.md records a 2026-08-22 brief asking an orchestrating agent to dispatch subagents for transport hardening, prompt expansion and output optimization, so parts of the code were developed with AI agents." ], "requirements_zh": [ "审阅时最新版本为 v3.0.5，于 2026-09-28 在 GitHub 发布，附带 wheel、源码包、SBOM 和 Sigstore 签名包，同时以 reversecore-mcp 之名发布到 PyPI，并提供 GHCR 镜像。所审阅的 main 提交（2026-10-17）比该标签多 147 个提交，但 pyproject.toml 仍声明 3.0.5。README 中的测试与覆盖率数字（3,069 个测试、88%）是写死的徽章，与 AGENTS.md 中较早的数字不一致。", "需要 Python 3.10–3.12，angr 只在 Python 3.12 上安装。本地安装需在 PATH 中提供 radare2，取证功能还需要 Volatility3、Scapy 和 Sleuth Kit；README 推荐使用已打包各引擎的 Docker 镜像，并提供 x86 与 Apple 芯片的 compose 配置。它列出的宿主为 Linux、macOS 或任意 Docker 环境，至少需要 4 核 CPU、8 GB 内存和 20 GB 存储；Redis 用于任务队列与结果缓存，SQLite 记忆数据库可跨会话保存笔记。", "二进制、内存镜像、PCAP 和磁盘镜像由分析者提供并挂载到工作目录，路径校验将工具访问限定在该目录内。动态分析工具的容器沙箱为可选项，默认关闭；除非设置 MCP_API_KEY，HTTP 模式只接受本机回环地址的客户端。部分工具和提示词面向漏洞利用与 PoC 开发。", "CI 运行代码检查、类型检查、Bandit、pip-audit 和 CodeQL，在 Python 3.10–3.12 上运行设有 85% 覆盖率门槛的单元测试，以及安全边界、集成和 MCP 客户端端到端测试，并检查 wheel 的构建与安装；ClusterFuzzLite 和每夜模糊测试另行运行。测试夹具包括带种子、概念验证输入和 ASan 日志的小型合成漏洞 C 程序，scratch/binaries 中只有一个 hello.c 源文件。", "采用 MIT 许可；所封装的引擎以及 Docker 镜像中的软件包保留各自的许可证。ORIGINAL_REQUEST.md 记录了 2026-08-22 交给编排智能体的任务说明，要求其分派子智能体完成传输层加固、提示词扩展和输出优化，可见部分代码是借助 AI 智能体开发的。" ], "targets": [ "binary" ], "url": "https://github.com/sjkim1127/Reversecore_MCP", "verification": "source-reviewed", "verified_at": "2026-10-27" }