Guosheng ToolboxBLOCK
为claude code+glm 添加上眼睛
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
English README | 中文说明
一个基于 Model Context Protocol (MCP) 的多模态 AI 工具箱,集成了智谱 GLM、Pollinations.AI、思源笔记和 Perplexity 网络搜索四大平台的强大能力。
🎯 为什么选择这个 MCP?
在使用 智谱 GLM 的 Claude Code 套餐时,我发现了一个限制:
- Lite 版本价格更低,但不支持 图像/视频理解 和 联网搜索;
- 如果要获得这些能力,就必须升级到 Pro(¥100/月) 或 Max(¥200/月);
- 但对很多开发者来说,仅仅为了这两个功能升级套餐,成本偏高。
我认为:
- 如果没有图像/视频理解,大模型就失去了"眼睛";
- 如果没有联网搜索,大模型就失去了"更新知识的能力";
- 这让 Lite 版本几乎成了一个"闭门造车"的模型,无法满足日常需求。
于是,我开发了这个 MCP 插件:
- 即使只用 GLM Lite 套餐,配合这个 MCP,也能为 Claude Code 装上眼睛(图像/视频理解)和 搜索引擎(联网搜索);
- 让 Lite 用户也能享受"最新知识 + 多模态理解"的体验;
- 这大大提升了 Lite 套餐的性价比,也是我认为本项目最大的价值。
🚀 功能特性
Bigmodel (智谱 GLM)
- 🖼️ 图片分析 - 使用 GLM-4.5V 模型进行图像识别和分析
- 🎬 视频分析 - 支持视频内容分析和理解
- 🎨 图片生成 - 使用 CogView-4 系列模型生成高质量图片
Pollinations.AI
- 🖼️ 图片生成 - 支持多种模型和参数配置
- 📝 文本生成 - 智能文本生成和对话
- 🔊 音频生成 - 文字转语音,支持多种声音
- 👁️ 图片分析 - OpenAI 兼容的视觉分析能力
思源笔记 (SiYuan)
- 📚 块级内容获取 - 获取思源笔记块的 Kramdown 源码
- 🔍 全文搜索 - 支持复杂查询语法的笔记搜索
- ✏️ 内容更新 - 直接更新思源笔记块内容
网络搜索
- 🔍 基础搜索 - 快速检索并返回原始搜索结果
- 🧠 智能搜索 - 使用 Sonar 模型生成答案并提供引用
- 🛡️ 安全过滤 - 内置安全机制防止恶意内容
- ⚡ 缓存加速 - 智能缓存提高重复查询响应速度
通用特性
- 🔧 灵活的环境变量配置
- 📝 完整的日志记录系统
- 🚀 轻量级,易于集成和部署
- 🛡️ TypeScript 类型安全
📦 安装
pnpm install
⚙️ 配置
API Key 设置
智谱 GLM API Key
访问 智谱AI开放平台 获取 API Key:
export GLM_API_KEY=your_glm_api_key_here
Perplexity API Key (网络搜索工具必需)
访问 Perplexity AI 获取 API Key:
export PERPLEXITY_API_KEY=your_perplexity_api_key_here
模型配置(可选)
# GLM 模型配置 GLM_IMAGE_MODEL=glm-4.5v GLM_VIDEO_MODEL=glm-4.5v GLM_GENERATION_MODEL=cogview-3-flash # 思源笔记配置(可选) SIYUAN_API_TOKEN=your_siyuan_token SIYUAN_API_BASE=http://127.0.0.1:6806 # 网络搜索配置(可选) WEB_SEARCH_CACHE_TTL=30 WEB_SEARCH_RATE_LIMIT=5 WEB_SEARCH_RATE_WINDOW_MS=60000 WEB_SEARCH_RETRY_AFTER_MS=1000 WEB_SEARCH_TIMEOUT_MS=10000
环境变量优先级
- 系统环境变量(推荐)
- 执行目录下的 `.en
4d8af89a274aOBSERVED · 2026-10-07Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.
claude mcp add mcpserver -- npx -y [email protected]
{
"mcpServers": {
"mcpserver": {
"command": "npx",
"args": [
"-y",
"[email protected]"
]
}
}
}Exposed tools (14)
13 read · 1 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
advanced_web_search | read | 在 web_search 的基础上,支持更复杂检索参数与结果整形;并可选启用 Sonar 直答(需要时才开)。 |
analyze_video | read | Analyze a video using GLM-4.5V model |
generate_image | read | Generate an image from text prompt using CogView-4 model |
get_block_kramdown | read | Get kramdown source code of one or multiple SiYuan note blocks |
pollinations_analyze_image | read | Analyze an image using Pollinations.AI Vision API |
pollinations_generate_audio | read | Generate speech audio from text using Pollinations.AI API |
pollinations_generate_image | read | Generate an image using Pollinations.AI API |
pollinations_generate_text | read | Generate text using Pollinations.AI API |
read_image | read | Analyze an image using GLM-4.5V model |
siyuan_database_query | read | |
siyuan_query_sql | read | |
siyuan_search_blocks | read | Search for blocks in SiYuan notes using full-text search |
update_block | write | Update a SiYuan note block with new content while preserving IAL metadata |
web_search | read | 快速检索并返回原始搜索结果(不做自动成文)。 |
Trust audit
BLOCKgrade D · trust 69/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | FAIL |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (5 observation(s))
- Network
- declared (7 observation(s))
- Shell
- none-observed
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (19)
'metadata.google.internal',
'169.254.169.254', // AWS metadata
import { getApiKey, encodeImageToBase64, callGLMApi } from '../../utils/helpers.js';import { createMessage, createErrorResponse, createSuccessResponse } from '../../utils/common.js';import { logger } from '../../utils/logger.js';import { getApiKey, generateImage } from '../../utils/helpers.js';import { createSuccessResponse, createErrorResponse } from '../../utils/common.js';SIYUAN_API_BASE=http://127.0.0.1:6806
- `SIYUAN_API_BASE`: 思源笔记 API 基础地址 (默认: http://127.0.0.1:6806)
SIYUAN_API_BASE=http://127.0.0.1:6806
- `SIYUAN_API_BASE`: 思源笔记 API 基础地址 (默认: http://127.0.0.1:6806)
SIYUAN_API_BASE=http://127.0.0.1:6806
@modelcontextprotocol/sdk, axios, dotenv, zod, @types/node, ts-node, typescript
Gates applied: no_behavioural_pass, no_license.
4d8af89a274afull audit observations/trust-audit/mcp-server/2234839__guosheng-toolbox.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-07 | 4d8af89a274a | BLOCK | D | 69 | first audit |
Questions
What is the Guosheng Toolbox MCP server?
为claude code+glm 添加上眼睛
What tools does Guosheng Toolbox expose?
14 in total: 13 read-only, 1 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is Guosheng Toolbox safe to connect to an agent?
No — not without reading the findings first. The audit graded it D (69/100) and found 2 critical or high issues in the source. Each one is listed on this page with the file and line it is on.
What credentials does Guosheng Toolbox need?
No credential environment variables were found in its source, so it appears to need none.
How does Guosheng Toolbox run?
It speaks stdio, so it runs as a local process your client starts. It is published on npm as mcpserver at 1.0.0.
How current is this page?
The grade is for one exact copy of the source (4d8af89a274a), read on 2026-10-07. The repository is watched and re-audited when it changes.