Guanlan
mcp
Warn
Health Warn
- License — License: MIT
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 5 GitHub stars
Code Pass
- Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
- Permissions — No dangerous permissions requested
Purpose
This is a Chinese internet research tool designed for AI agents. It routes queries to appropriate sources, reads web pages, observes trending topics, and compiles structured evidence packages for the agent to process.
Security Assessment
Overall risk: Low. The tool makes external network requests to perform web searches, fetch RSS feeds, and read web pages, which is its intended function. The code scan of 12 files found no hardcoded secrets, dangerous patterns, or malicious code. It does not request dangerous system permissions. The README explicitly instructs agents not to read browser cookies, trigger login authorizations, or request keychain access, indicating a strong security-first design philosophy by the developer.
Quality Assessment
The project uses the standard MIT license and was actively updated recently. However, it currently has very low visibility with only 5 GitHub stars, meaning it has not been broadly reviewed by the open-source community. The documentation is thorough, including a built-in doctor command for verifying installations and preventing path drift, which reflects good development practices. It is clearly labeled as being in the "alpha" stage, so users should expect potential instability or breaking changes.
Verdict
Use with caution — the tool appears safe and well-designed, but its alpha status and lack of widespread community adoption mean it should be tested in isolated environments before production use.
This is a Chinese internet research tool designed for AI agents. It routes queries to appropriate sources, reads web pages, observes trending topics, and compiles structured evidence packages for the agent to process.
Security Assessment
Overall risk: Low. The tool makes external network requests to perform web searches, fetch RSS feeds, and read web pages, which is its intended function. The code scan of 12 files found no hardcoded secrets, dangerous patterns, or malicious code. It does not request dangerous system permissions. The README explicitly instructs agents not to read browser cookies, trigger login authorizations, or request keychain access, indicating a strong security-first design philosophy by the developer.
Quality Assessment
The project uses the standard MIT license and was actively updated recently. However, it currently has very low visibility with only 5 GitHub stars, meaning it has not been broadly reviewed by the open-source community. The documentation is thorough, including a built-in doctor command for verifying installations and preventing path drift, which reflects good development practices. It is clearly labeled as being in the "alpha" stage, so users should expect potential instability or breaking changes.
Verdict
Use with caution — the tool appears safe and well-designed, but its alpha status and lack of widespread community adoption mean it should be tested in isolated environments before production use.
观澜 / Guanlan：AI Agent 的中文互联网研究、阅读与信源路由工具。
README.md
观澜 / Guanlan
让 AI Agent 看懂中文互联网
临流观势，循源取义。观澜是面向 AI Agent 的中文互联网研究工具：会路由信源、阅读网页、观察热榜、整理证据。
30 秒 TL;DR · 安装 · 安全边界 · 文档
30 秒 TL;DR
普通搜索让 Agent 看见网页，观澜让 Agent 看懂中文互联网。
观澜不是“又一个搜索 CLI”，而是给 AI Agent 用的中文互联网研究底座：它会先判断该去哪里找，再把网页、热榜、RSS、官方源、垂类媒体和社区样本整理成可追溯的证据上下文。
如果你只想先用起来：
uv tool install --force --upgrade guanlan
guanlan version
guanlan search "人工智能 政策 最新" --profile china --limit 80
最稳能力
- 中文网页搜索：公开网页搜索、去重、排序、来源标注，默认给 Agent 足够大的候选池。
- 信源路由：政策看官方和党央媒，口碑看社区样本，技术看开发者与 RSS，产业看垂类媒体。
- 网页阅读：把网页转成 Markdown，并提供正文质量、噪声、fallback 和 trace。
- 热榜观察：用
hotnews和feeds帮 Agent 观察“今天中文互联网在涌动什么”。 - 证据包输出：
research、compare、timeline、dossier、archive、--format context面向 Agent 继续推理。
典型命令
# 1. 搜中文网页，默认建议给 Agent 足够候选池
guanlan search "新质生产力 政策 原文" --profile china --limit 80 --trace
# 2. 读取网页正文，检查是否干净
guanlan read "https://example.com/article" --quality-report --trace
# 3. 做一份 Agent 可直接使用的研究证据包
guanlan research "某产品 用户评价 值不值得买" --preset reputation --advisor --format context
# 4. 看今日中文互联网水势
guanlan hotnews today --limit 80 --trends
# 5. 不确定去哪搜时，先看信源路由
guanlan route "AI 产业政策 地方 补贴" --json
更多命令见 完整使用手册 和 Agent 使用指南。
安装
推荐任选一种方式。安装或升级后，先确认版本和路径，避免 Agent 调到旧版。
Homebrew
brew update
brew tap shenyangs/tap
brew reinstall shenyangs/tap/guanlan
uv
uv tool install --force --upgrade guanlan
pipx
pipx install --force guanlan
验证
guanlan version
guanlan doctor --install-check
guanlan status
看到 观澜 / Guanlan v0.5.10，并且安装检查没有版本/路径漂移，就说明基础部署成功。
给 Agent 复制的安装指令
请帮我安装观澜 CLI，并验证基础功能可用。
优先使用：
uv tool install --force --upgrade guanlan
安装完成后运行：
guanlan version
guanlan doctor --install-check
guanlan status
如果 guanlan version 不是 README 标注的当前版本，请不要继续配置 MCP 或可选渠道，先排查 PATH 和安装来源。
安全要求：不要读取浏览器 Cookie，不要触发登录授权，不要请求钥匙串权限。
安全边界
观澜默认只读、低扰、明源。
- 默认不读取浏览器 Cookie、钥匙串、登录态或私密配置。
- 浏览器辅助补证只读取用户授权后的目标页可见内容；Cookie 读取必须另行说明平台、用途、风险和只读范围，并获得单独明确授权。
- 不自动点赞、评论、关注、发帖或发送消息。
- 高风险平台能力保持
best-effort/opt-in/experimental口径，不包装成稳定端到端能力。 doctor默认不做深度授权探测；需要认证检查必须显式运行相关命令。- Archive、Wiki、RAG 导出都基于本地已归档资料，不代表全网事实。
- 真实网络超时应被视为网络/上游证据，不应直接解释为“没有结果”。
文档入口
| 你想做什么 | 去哪里看 |
|---|---|
| 完整命令和长说明 | 完整使用手册 |
| Agent 怎么正确调用观澜 | Agent 使用指南 |
| 安装、升级和路径排查 | 安装指南 / 更新指南 / 故障排查 |
| 本地模型联网 | Local LLM 指南 |
| 精选信源包 | Source Packs |
| 热点目录接入 | 热点目录接入 |
| 输出字段稳定契约 | Contract |
| 中文互联网设计依据 | China-Aware Web Rationale |
| 发版和质量闸门 | Release Automation / Quality Test Plan |
| 路线图 | Roadmap |
项目定位
观澜当前仍是 Alpha 阶段的 CLI-first 工具。最稳路径是公开搜索、网页阅读、热榜观察、研究证据包和本地归档；部分平台能力依赖公开页面、外部后端、网络环境或显式授权。
临流观势，循源取义。观澜让 Agent 在进入中文互联网之前，先学会“观”。
License
MIT License. See LICENSE for details.
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found