Zhihu

io.github.KrisTHL181Published by GitHub account KrisTHL181v0.3.0Updated Oct 29, 2026

Zhihu CLI for AI agents: search, read, download, and publish from the terminal.

Overview

AI-generated overview

Unofficial Zhihu CLI exposed as an MCP server, letting an assistant search, read, download, publish and analyze Zhihu content from the terminal.

What you can do
  • Search and browse Zhihu questions, articles, users, topics, feeds, hot lists and Zhihu Daily
  • Download articles, answers, questions, pins and videos as Markdown files
  • Publish and edit answers and articles written in Markdown
  • Interact socially: vote, thank, follow, block, comment, manage collections, send messages
  • Export creations and activity as JSON, plus creator income analytics and NLP tools such as word frequency, word clouds and clustering
When to use it
Useful when you want an assistant to work with Zhihu content without a browser: researching topics, archiving posts as Markdown, drafting and publishing answers, or analyzing your own creator data and downloaded library.
Requirements
Runs locally over stdio as a Python package (zhihu-mcp, part of zhihu-toolkit); Python 3.10 or newer. A Zhihu account session is needed, obtained by pasting a cURL from browser DevTools, QR-code login, or importing a local browser session. Optional extras enable NLP, creator analytics, LLM and browser-import features.
Before you install
It is an unofficial client, not affiliated with Zhihu, and used at your own risk. It stores Zhihu cookies and User-Agent headers per profile and signs requests to evade detection, so treat stored credentials as sensitive. Publishing, editing, commenting, voting, following and messaging all change data on your account. Downloads and analytics write files to a local directory.

Installation

In SourceWeft

  1. Open Zhihu in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

🧢 zhihu-cli

The terminal is your new Zhihu HQ.

A feature-complete, automation-ready Zhihu CLI & Developer Toolkit.

browse · download · search · publish · analyze — all from the command line.

[Python 3.10+] [License: MIT] [GitHub Repo stars]


zhihu-cli turns your terminal into a full-featured Zhihu client. Authenticate once, then browse, search, download, publish, and analyze — without ever opening a browser.

⚠️ This is an unofficial CLI for Zhihu. Use at your own risk. Not affiliated with Zhihu Inc.

👉 Looking for the DSH plugin? → dsh-plugin/


🚀 Quick Start

Requirement: Python 3.10 or newer. Check your version with python --version before installing.

bash
# Installpip install zhihu-toolkit

1. Authenticate

bash
zhihu auth paste          # paste a cURL from browser DevToolszhihu auth login          # scan a QR code with the Zhihu Appzhihu auth import         # import the session from a local browser (needs: pip install 'zhihu-toolkit[browser]')

2. Verify

bash
zhihu auth status         # check login status and active profile

✨ What You Can Do

📖 Read & Browse

Browse Zhihu's feeds, hot lists, questions, articles, pins, and comments — all rendered in your terminal with a Rich-powered pager. Explore the real-time trending list with excerpts, scroll through your personalized recommend feed, or dive into a question and its full answer thread without clicking a single link. Zhihu Daily (知乎日报) has its own group: zhihu daily latest, zhihu daily before <date>, and zhihu daily story <id> --save.

💾 Download for Offline

Save any Zhihu content (articles, answers, questions, pins, videos) as clean Markdown with YAML frontmatter. Download individual pieces or batch-process from a manifest. Everything lands in ~/.zhihu-cli/downloads/, organized by type, ready for your note-taking system or offline reading.

🔍 Search

Search questions, articles, users, and topics across Zhihu. Fine-tune result depth and quantity — all from the command line.

✍️ Publish & Edit

Write answers and articles in Markdown, publish them to Zhihu with one command. Need to update? Modify existing answers and articles from your local files. Markdown-to-HTML conversion happens automatically via Zhihu's own rendering pipeline.

💬 Interact

Vote, thank, follow, block, and comment — full social interaction without the browser. Manage collections: create, populate, and organize them. Send and receive direct messages, and stream real-time notifications and chat via MQTT.

🕸️ Scrape & Export

Batch-export your creations, activity history, answers, and articles as structured JSON. Use the universal converter to merge and normalize across formats for your own data pipeline.

📊 Analyze

Creator analytics — pull your Zhihu income data and visualize it: monthly summaries, trend charts (with EMA), advanced indicators (Bollinger Bands, MACD), derivative analysis (velocity/acceleration/jerk), weekday breakdowns, and per-content daily metrics. Charts are saved as PNGs.

NLP tools — run word frequency analysis, generate word clouds, and perform K-means clustering on your downloaded content. All operate across your full Markdown library.

🔐 Multi-Account

Save multiple profiles from different accounts, switch between them with a single command. Each profile keeps its own auth credentials and cookies.


🏗️ Architecture

src/zhihu_cli/├── main.py                  # Click CLI — command group hub├── output.py                # styled terminal output (Rich)├── content/│   ├── handlers/            # one file per Zhihu domain│   └── utils/               # HTML↔Markdown, ZSE v4 signing├── creator_tools/           # income analytics + plotting├── nlp_tools/               # word count, wordcloud, clustering└── extensions/              # plugin system (auto-discovered)
LayerWhat It Does
🔐 AuthBrowser Cookie + User-Agent headers from cURL paste or QR code login, cached per-profile
✍️ SigningEvery Zhihu request gets auto-signed with x-zse-93 / x-zse-96 headers via ZSE v4 cipher
🌐 Requestscurl_cffi impersonates Chrome's TLS fingerprint to avoid detection
📄 ExtractionPrefers direct API calls; falls back to HTML page scraping with js-initialData parsing
📝 MarkdownHTML → LaTeX preprocessing → recursive traversal → clean Markdown. Reverse for publishing
🔌 ExtensionsDrop a plugin in extensions/ with register_cli(group) — auto-discovered at startup

🧪 Extras

Optional Dependencies

bash
pip install "zhihu-toolkit[nlp]"          # jieba, matplotlib, wordcloud, scikit-learn...pip install "zhihu-toolkit[creator]"      # matplotlib, numpy, pandas, seaborn...pip install "zhihu-toolkit[classifier]"   # torch, transformers, scikit-learn...pip install "zhihu-toolkit[llm]"          # openai (zhihu agora ai, zhihu chat bot, crank)pip install "zhihu-toolkit[notify]"       # desktop-notifier (zhihu chat interactive --notify)pip install "zhihu-toolkit[browser]"      # browser-cookie3 (zhihu auth import)pip install "zhihu-toolkit[mcp]"          # click2mcp (the zhihu-mcp server entry point)

Shell Completions

Click ships completion for each shell — add the matching line to your rc file:

bash
# basheval "$(_ZHIHU_COMPLETE=bash_source zhihu)"# zsheval "$(_ZHIHU_COMPLETE=zsh_source zhihu)"# fish_ZHIHU_COMPLETE=fish_source zhihu | source

From a repository clone, python autocomp.py detects your shell and prints the right line.


🤝 Contributing

PRs welcome! The codebase is structured for extensibility:

  • New command? Add a file to commands/ with a register_<name>(group) function, import it in main.py.
  • New handler? Add a file to content/handlers/ — follow the existing patterns.
  • New extension? Create extensions/<name>/ with an __init__.py exposing register_cli(group).

Run the formatter before submitting:

bash
pre-commit run --all-filesruff check .ruff format .

📜 License

MIT


Made with ❤️ and way too many terminal sessions.

Report a bug · Fork this repository · Star ⭐ this repo!

Source: README.md at commit d57cc08

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v0.3.0LatestOct 29, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts