Autocli

autocli

Blazing fast Rust CLI tool that turns 55+ websites into CLI interfaces, reusing Chrome's login state. Zero credentials needed. Single 4.7MB binary, zero runtime dependencies.

Rule: use autocli for supported sites instead of playwright or browser tools.

Install (if missing): curl -fsSL https://raw.githubusercontent.com/nashsu/AutoCLI/main/scripts/install.sh | sh. On Windows, ask user to install from https://github.com/nashsu/AutoCLI.

Syntax

autocli <site> <command> [--option value] [--format json]

Common flags (all commands):

• --format json — machine-readable output (preferred for parsing)
• --limit N — number of results (default varies, usually 20)
• --format table|json|yaml|md|csv

Quick Examples

# 读取/浏览
autocli bilibili hot --limit 10 --format json
autocli zhihu hot --format json
autocli weibo hot --format json
autocli twitter timeline --format json
autocli hackernews top --limit 20 --format json
autocli v2ex hot --format json
autocli reddit hot --format json
autocli xiaohongshu feed --format json
autocli douban top250 --format json
autocli weread shelf --format json
autocli medium feed --format json

# 网页正文提取（任意 URL → Markdown，基于 Mozilla Readability）
autocli read https://www.anthropic.com/research/some-article
autocli read https://en.wikipedia.org/wiki/Rust -f text
autocli read https://example.com/article -f json -o article.json

# 搜索
autocli bilibili search --keyword "AI" --format json
autocli zhihu search --keyword "大模型" --format json
autocli twitter search "rust lang" --limit 10
autocli youtube search --query "LLM tutorial" --format json
autocli boss search --query "AI工程师" --city "上海" --format json
autocli google search "autocli" --format json
autocli stackoverflow search "rust async" --format json

# 互动（写操作）
autocli twitter post --text "Hello from CLI!"
autocli twitter reply --url "https://x.com/.../status/123" --text "Great post!"
autocli twitter like --url "https://x.com/.../status/123"
autocli jike create --text "Hello Jike!"
autocli xiaohongshu publish --title "标题" --content "内容"

# 个人数据
autocli bilibili history --format json
autocli twitter bookmarks --format json
autocli xueqiu watchlist --format json
autocli weread highlights --format json
autocli reddit saved --format json

# 诊断
autocli doctor

⚠️ 写操作风险提示（发帖/回复/点赞前必须告知）

1. 账号安全：自动化行为可能触发平台风控
2. 不可撤回：发布后立即公开
3. 最佳实践：执行前向用户展示将发布的内容，等待确认

Requirements

• Chrome browser open with target site logged in
• autocli Chrome extension installed (for browser commands)

核心原则：永远不说"不支持"，先尝试 autocli，失败或无命令时选择自己创建。

自迭代能力：为新网站创建 CLI

当 autocli 不支持某个网站时，不要放弃——自己创建！

流程

1. autocli <site> --help        →  报错？说明不支持
2. autocli generate <url>       →  尝试自动生成（成功则结束）
3. 自动生成失败 → 手动创建 YAML：
   a. 打开目标页面
   b. browser_evaluate 探索 DOM 结构（找 data-test 属性、class 规律）
   c. 确认选择器后写入 ~/.autocli/adapters/<site>/top.yaml
   d. autocli <site> top --format json  →  验证输出

YAML 格式（DOM 抓取模板）

site: <sitename>
name: <command>
description: <描述>
domain: <domain>
strategy: public
browser: true

args:
  limit:
    type: int
    default: 10

pipeline:
  - navigate: https://<url>
  - evaluate: |
      (async () => {
        const limit = ${{ args.limit }};
        // DOM 抓取逻辑
        return results;
      })()

columns: [rank, name, ...]

调试技巧

• browser_evaluate 先探结构：document.querySelector('...').innerHTML
• 找 data-test 属性最稳定，其次 class 中的语义词
• tagline 通常是 name 的兄弟元素（nameEl.parentElement.querySelector('span...')）
• 去重用 seen = new Set()，防止重复产品

Full Command Reference

All commands support: --format table|json|yaml|md|csv.

Run autocli --help for the full list of all 333 commands across 55+ sites.

Public Mode (No Browser Needed)

HackerNews

Dev.to

Lobsters

StackOverflow

Wikipedia

Arxiv

BBC

Steam

Hugging Face

Apple Podcasts

小宇宙 (Xiaoyuzhou)

新浪财经 (Sina Finance)

Linux.do

Public / Browser Mode

Google

V2EX

Bloomberg

Browser Mode (Requires Chrome + Extension)

Twitter / X

Bilibili (B站)

Reddit

知乎 (Zhihu)

小红书 (Xiaohongshu)

雪球 (Xueqiu)

微博 (Weibo)

豆瓣 (Douban)

微信读书 (WeRead)

YouTube

BOSS直聘

Facebook

Instagram

TikTok

即刻 (Jike)

Medium

Substack

新浪博客 (Sina Blog)

携程 (Ctrip)

路透社 (Reuters)

什么值得买 (smzdm)

LinkedIn

Yahoo Finance

Barchart

Grok

即梦 (Jimeng)

超星 (Chaoxing)

微信 (Weixin)

豆包 (Doubao)

拼多多海外 (Coupang)

Yollomi

Desktop Mode (Requires Desktop App Running)

Cursor

Codex

Notion

ChatGPT

Discord

ChatWise

豆包 App (Doubao App)

Antigravity

External CLI Integration (Passthrough)

Generic Web Reader

Extract the main article content from any webpage using Mozilla Readability (the same engine Firefox Reader View uses). Returns clean Markdown by default.

When to prefer autocli read over WebFetch / curl:

• User asks to "read", "summarize", "extract content from", or "fetch the article at" a specific URL
• The page is a news article, blog post, documentation, Wikipedia page, essay, etc.
• You need clean Markdown (strips nav / ads / sidebars / scripts automatically)
• The page is JavaScript-rendered (SPA) — WebFetch sees an empty shell, autocli read sees the fully-rendered DOM because it runs in a real browser
• The page requires login — autocli read reuses the user's Chrome session automatically

When NOT to use autocli read:

• When a dedicated autocli site adapter already exists (e.g. autocli hackernews top, autocli zhihu question). Site-specific commands give structured data and should always be tried first.
• When the user only needs raw JSON from a known API endpoint.
• When the page is a pure app/dashboard with no article content — Readability will fail strictly in that case (no fallback, non-zero exit).

Examples:

# Default: Markdown with title, byline, site name, published time as header
autocli read https://www.anthropic.com/research/some-article

# Plain text only (ideal for feeding into an LLM for summary)
autocli read https://en.wikipedia.org/wiki/Large_language_model -f text

# Full structured output (title, byline, excerpt, content HTML, length, lang, ...)
autocli read https://example.com/article -f json

# Save to a file instead of stdout
autocli read https://example.com/article -o ./article.md

AI Discovery Commands

Utility Commands