中文 · Русский (RU)
MCP 服务器，用于 Pentool —— 本地 AI 助手，助力 Web 渗透测试
一个自包含的 stdio JSON-RPC 2.0 服务器，为任何 MCP 客户端（尤其是 Pentool）带来 AI 能力（BYO-LLM：选择检查项、绕过 WAF、发现非明显端点）。通过 uv 一条命令安装。
MCP 是连接 LLM 模型与工具的开放协议。pentool-mcp-server 将 Pentool 的 MCP 层封装为可复用的 PyPI 包：本地 stdio 附加，无网络，无端口，秒级安装，无重量级依赖。
Related MCP server: pentestMCP
此服务器是 Pentool 的 MCP 组件，Pentool 是一个专业的 Web 安全测试终端（兼容 Burp 的代理、扫描器、爬虫、入侵器——全部在 TUI 中）。它为 Pentool 的 AI 提供能力：
🎯 针对具体目标选择相关的 扫描检查项；
🛡 WAF 绕过 / 载荷建议；
🕷 在爬虫爬取过程中发现 非明显端点。
尝试完整栈：
uv tool install "pentool[ai]"→pentool ai setup→pentool。
# Install uv (if not present): https://docs.astral.sh/uv/
curl -LsSf https://astral.sh/uv/install.sh | sh
# Standalone — just the MCP server
uv tool install pentool-mcp-server
pentool-mcp-server --version
# Or together with Pentool and its AI extras
uv tool install "pentool[ai]"健康检查（stdio）：
echo '{"jsonrpc":"2.0","id":1,"method":"ping"}' | pentool-mcp-server
# → {"jsonrpc": "2.0", "id": 1, "result": {"status": "ok"}}uv 是标准安装路径——隔离环境，与 Pentool 自身的安装方式完全一致。它使服务器远离系统 Python。
服务器不开放任何端口，作为本地子进程运行——在任何主机上均安全。使用 uv 可使其远离系统 Python：
uv tool install pentool-mcp-server完整用户指南——docs/GUIDE.md。俄语版——README.ru。
工具 | 描述 |
| MCP 协议握手 |
| 列出可用工具 |
| 为任务生成 LLM 回答（ |
| 就绪检查（模型是否已安装） |
| 指向 GGUF 模型路径 |
| 进程存活检查 |
默认无网络。 pentool-mcp-server 仅在本地进程内监听 stdio（标准输入→标准输出）。外部访问不可能——既不能从其他进程，也不能从其他主机。
未经认证，请勿通过 TCP/0.0.0.0 暴露。 如果在外部网络端口上运行，任何能够写入标准输入/端口的人都可以调用 tools/call generate（资源消耗 + 向 LLM 发送目标数据）。MCP 没有内置认证——请绑定到 127.0.0.1 并通过防火墙进行防护。
目标隐私。 数据（URL、载荷）仅在连接外部 LLM 提供商时才会离开主机；使用本地 GGUF 模型时，流量保留在机器上。
更多——docs/GUIDE.md。
stdio JSON-RPC 2.0 服务器（initialize / tools/list / tools/call / ping）
generate、health、configure 工具
在 model.py 中实现真正的 LLM 运行器（llama-cpp-python）
可选的 TCP 模式，带 HMAC 认证
AGPL-3.0 —— 与 Pentool 相同的许可证。
Принять параметры конфигурации сервера (напр. путь модели).
| Name | Required | Description | Default |
|---|---|---|---|
| model | No | Путь к GGUF-файлу модели |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description must fully disclose behavior. It only says 'accept parameters' without explaining side effects, persistence, validation, or restart requirements. This is insufficient for a configuration tool.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, front-loaded sentence with no wasted words. It is concise and to the point, though it sacrifices valuable behavioral detail for brevity.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has one optional parameter, no output schema, and no annotations, the description still leaves key context missing: what configuration does, whether it is required for sibling tools, and what happens after invocation. It is too sparse to be fully complete.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, and the description's mention of 'model path' merely mirrors the schema's own description for the 'model' parameter. No additional meaning is added, so the baseline score of 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states it accepts server configuration parameters with an example (model path), which clearly identifies the tool's purpose. It is distinct from siblings 'generate' and 'health', though the verb 'accept' is somewhat passive and the exact effect is implied rather than explicit.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance is provided on when to use this tool versus alternatives, nor are prerequisites or ordering requirements (e.g., configure before generate) mentioned. The description implies usage context but gives no explicit when/when-not guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Сгенерировать ответ LLM по задаче. payload — строка запроса (данные цели), model — путь к GGUF-файлу модели (необязательно). При отсутствии модели вернёт ошибку setup required.
| Name | Required | Description | Default |
|---|---|---|---|
| task | Yes | Имя задачи (напр. choose_checks, crawl_endpoints) | |
| model | No | Путь к GGUF-файлу модели | |
| payload | Yes | Данные цели / контекст (JSON-строка) |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It discloses that the model parameter is optional and that omitting it causes an error ('setup required'), which is useful behavioral insight. However, it does not mention whether the tool is read-only or has side effects (e.g., logging), or any rate limits or auth requirements. The behavior is partially transparent.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is very concise at two sentences, front-loading the main purpose. Every phrase adds value, though the error clause could be more specific (e.g., 'returns an error' instead of 'вернёт ошибку'). No extraneous text, well-structured for quick parsing.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given no output schema, the description should clarify what the tool returns (e.g., generated response text). It does not, leaving a gap. With only 3 params all documented in schema, the description is adequate for basic use but incomplete for understanding return behavior or error responses beyond the 'setup required' case.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, and the description adds some value by explaining that model is optional and payload is the query data. It does not elaborate on 'task' beyond the example values in schema, and does not clarify the format of payload (JSON string) beyond the schema. As coverage is high, baseline 3 is appropriate given the minimal added context.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states the tool generates an LLM response for a task, specifying the payload as the query string and model as an optional GGUF file path. It clearly identifies the verb (генерировать) and resource (ответ LLM по задаче), and distinguishes from siblings like 'health' and 'configure' by its focus on generation rather than status checks or setup.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage when an LLM response is needed for a task, and mentions that missing model leads to a 'setup required' error. However, it does not explicitly state when to use this vs. 'configure' (e.g., to set the model first) or when not to use it (e.g., if no model is configured). The sibling tools are named but no direct comparison is made.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Проверка готовности: вернёт статус, установлена ли модель.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden. It indicates the tool is read-only (returns status) and non-destructive, but does not disclose error behavior, return format details, or potential side effects. The description is minimally transparent.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, short sentence that is front-loaded and contains no redundant information. Every word earns its place.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's simplicity, the description is mostly adequate for a health check. However, it lacks specification of the return value format (e.g., boolean, JSON) and does not mention any edge cases. With no output schema, more detail would improve completeness.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has zero parameters and schema coverage is 100% (trivially). Per the baseline rule for 0 parameters, the description does not need to add parameter semantics, and it appropriately omits any.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool performs a readiness check and returns whether the model is installed. The verb 'Проверка готовности' (readiness check) and resource 'model' are specific. However, it does not explicitly differentiate from sibling tools 'generate' and 'configure', though the purpose is distinct enough.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance is provided on when to use this tool versus alternatives. There is no mention of preconditions, context, or scenarios where 'health' is appropriate compared to 'generate' or 'configure'.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.0configuregeneratehealthScored across 3 tools
Each tool has a clearly distinct purpose: 'generate' for LLM inference, 'health' for readiness checking, and 'configure' for setting parameters. There is no overlap or ambiguity between them.
All tool names are simple, lowercase verbs that accurately describe their action ('generate', 'health', 'configure'). The pattern is uniform and predictable.
With 3 tools, the count is appropriate for a minimal MCP server focused on LLM inference. It could potentially benefit from a status tool or a list-models tool, but the current set is sufficient for its core purpose.
The tool surface covers basic lifecycle: configure (setup), health (readiness), and generate (infer). However, there is no tool for cancelling generation, listing available models, or inspecting current configuration, which are notable gaps for a production server.
MCP server for Pentest-Tools.com: run scans, manage findings and reports via your preffered LLM.
Security scanner for MCP servers. Detect vulnerabilities, prompt injection, and tool poisoning.
Security-first WordPress MCP server. 129 tools for Claude, ChatGPT, Gemini. Free on wp.org.
Self-hosted MCP server: 26 deterministic dev, security, and EVM tools.
