Agent-Reach

1 align="center">👁️ Agent Reach</h1> <p align="center"> <strong>Equip Your AI Agent with One-Click Internet Access</strong> </p> <p align="center"> The most stable connection method, selected, installed, and tested for you - no need to worry about upgrades. </p> <p align="center"> <a href="LICENSE"><img://img.shields.io/badge/License-MIT-blue.svg?style=for-the-badge" alt="MITa> <a href="https://www.python.org/"><img src="img.shields.io/badge/Python-3.10+-green.svg?style=for-the-badge&logo=python&logoColor=white" alt="Python 3.10+"></a> <a href="https://github.com/Panniantong/agent-reach/stargazers"><img src="https://img.shields.io/github/stanniantong/agent-reach?style=for-the-badge" alt="GitHub Stars"></a> <a href="https://trendshift.io/repositories/24387"><img src="https://trendshift.io/api/badge/repositories87" alt="Trendshift GitHub Trending #1 Repository of the Day"></a> </p> <pcenter"> <a href="#快速上手">Quick Start> · <adocs/README_en.md">a> · <a href="docs/README_ja日本語</a> · <a href="docsko.md">한국어> · <a href="#支持的平台">Supporteda> · <="#设计理念">Design Philosophy</a> </p> Do We Need? AI Agents can already write code, modify and manage projects - but when you ask them to find something online, they struggle: - 📺 "Can you check out this YouTube tutorial?" → **Can't access**, no subtitles available - 🐦 "Can you search Twitter for reviews of this product?" → **Can't search**, Twitter API requires payment - 📖 if anyone on Reddit has encountered a similar bug" → blocked**, server IP- 📕 the reputation of this Xiaohongshu" → **Can't access**, - 📺 " Bilibili → **Can't access**, tools blocked by B security - 🔍 "Search for the latest LLM online" → **No good**, either paid or poor quality - 🌐 this webpage says" → **Returns a lot of HTML tags**, hard to read - 📦 "What does this GitHub repository do? What are the issues?" → Can be used, but authentication configuration is complicated - 📡 "Subscribe to these RSS feeds and notify me of updates" → Requires custom library and code ** are not difficult, but require configuration and setup** Each platform has its own requirements - paid APIs, firewalls to bypass logins, and data. You have to go through each one, install and configure settings. Just enabling an Agent to read Twitter requires a lot of effort. **Agent Reach simplifies this to one line:** ``` Help me install Agent Reach:raw.githubusercontent.com/Panniantong/agent-re/docs/install.md ``` Copy this to your Agent, and in a few minutes, read Twitter, search Reddit, watch YouTube, and browse Xiaohongshu. **Already installed? Updating is also one line:** ``` Help me update Agent Reach: https://.com/Panniantong/agent-reach/main/docs/update.md ``` > ⭐ **Star this project**, and we will continuously track changes to each platform and add new channels. You don't need to monitor - if a platform blocks us, we will fix it, and if there are new channels, we will add them. ### ✅ Before You Use | | | |---|---| | 💰 **Completely free** | All tools are open-source, and all APIs are free. The only potential cost is server proxy ($1/month), not required for local| 🔒 **Privacy and security** | Cookies are, not uploaded or shared. Code is completely open-source be reviewed at any time | | 🔄 **Continuous updates** | Each platform has a primary and backup channel. If one channel fails, we switch to the next, and you won't notice (e.g., yt-dlp was blocked by Bilibili security in June 2026, but we switched to bili-cli with no user action) | | 🤖 **Compatible with all Agents Claude Code, OpenClaw, Cursor, Windsurf... any Agent run shell commands Agent Reach | | Self-diagnostics**agent-reach doctor` provides a one-line check which channels are working, which are not, and how to fix them | Supported Platforms | Platform | Ready to use | Configurable | How to configure | |------|---------|-----------|-------| 🌐 **Webpage** | Read any webpage | — | No configuration required | | 📺 **YouTube** | Extract subtitles + search videos | — | No configuration required | | 📡 **RSS** | Read any RSS/Atom feed | — | No configuration required | | 🔍 **Full-text search** | — | Semantic search of the entire web | Automatic configuration (MCP access, free without API key) | 📦 **GitHub** | Read public repositories + search | Private repositories, issues/PR, Fork | Tell Agent "Help me log in to GitHub" | | 🐦 **Twitter/X** | Read single tweets | Search tweets, browse timeline, read long articles | Tell Agent "Help me configure Twitter" | | 📺 **Bilibili** | Search + video details (bili-cli, no login required) | Subtitles (OpenCLI) | Tell Agent "Help me configure Bilibili" | | 📖 **Reddit** | — (no zero-configuration path: anonymous interface blocked) | Search + read posts and comments | Install OpenCLI on desktop and log in with browser; or rdt-cli + Cookie | | 📘 **Facebook** | — | Search, homepage, feed, group list | Install OpenCLI on desktop (reuse Chrome login state) | | 📷 **Instagram** | — | User search, profile, recent posts, explore | Install OpenCLI on desktop (reuse Chrome login state) | | 📕 **Xiaohongshu** | — | Search, read, comments | Install OpenCLI on desktop (browse Xiaohongshu to use); server uses xiaohongshu-mcp to scan QR code | | 💼 **LinkedIn** | Jina Reader reads public pages | Profile details, company pages, job search | Tell Agent "Help me configure LinkedIn" | | 💻 **V2EX** | Popular posts, node posts, post details + replies, user information | — | No configuration required | | 📈ueqiu** | Stock quotes, search stocks, popular posts, popular stock rankings | — | Tell Agent "Help me configure Xueqiu" | | 🎙️ **Xiaoyuzhou Podcast** | — | Podcast audio transcription (Whisper transcription, free API key) | Tell Agent "Help me configure Xiaoyuzhou Podcast" | > **Don't know how to configure? Don't check the documentation.** Just tell the Agent "Help me configure XXX", and it will guide you through the process. > > 🍪 For platforms that require cookies/login state (Twitter, Xiaohongshu, Reddit, Facebook, Instagram, etc.), users are prioritized to log in with their own browser. OpenCLI reuses Chrome's login state; traditional CLI requires Cookie-Editor to export cookies. > > 🔒 Cookies are stored locally, not uploaded or shared. Code is completely open-source and can be reviewed at any time. > 💻 Local computers do not require a proxy. Proxies are only required when deploying on servers (~$1/month). --- ## Quick Start > ⚠️ **OpenClaw users, please ensure exec permissions are enabled** > > Agent Reach relies on the Agent commands (`pip `mcporter`,`, etc.). IfClaw uses themessaging` tool configuration, the Agent will not to execute commands. **Enable exec permissions before installation**: > > ```bash config set tools" > ``` > or set `"tools": { " "coding" }` in `~/.openclaw/openclaw.json`. > Restart the Gateway (`open restart`) and start a new conversation. Copy this Agent (Claude Code, OpenClaw, Cursor, etc.): ``` Help me install Agent Reach: https://raw.githubusercontent.com/Panniantong/agent-re/install.md ``` That's it. The Agent will complete the rest of the setup. > 🔄 **Already installed?** Updating is also: > ``` > Help me update Agent Reach: https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/update.md > ``` > 🛡️ **Concerned about security?** Use safe mode - will not automatically install system packages, only lists what is needed: > ``` > Help me install (safe mode): https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md > Use parameter during installation ``` <details> <summary>What does itClick to expand)</summary> 1. **Install CLI tools** — `pip install` installs the `agent-reach` command line (includes yt-dlp,) 2. **Install system infrastructure** — automatically detects and installs Node.js, gh CLI, mcporter 3. **Configure search engine** — accesses Exa through MCP (free, no API) 4. **** — determines if it's a local computer or server, provides configuration suggestions 5. **Register SKILL.md** — installs usage guidelines Agent's skills directory; later, when the Agent encounters "full-text survey", "search Twitter", "watch video", it will automatically know which tool to use 6. **Ask if you want more** — by only 6 zero are activated; like Xiaohongshu, Twitter, Reddit, Facebook, and Instagram that require login state, the Agent will list a menu and ask which ones you want to enable After installation, `agent-reach provides a one-line command to check the status of each channel and suggest fixes. </details> --- ## Ready to use No configuration required, just tell the- "Check this link" → `curl https://r.jina.ai/URL` reads any webpage - "What does this GitHub repository do?" → `gh repo view owner/repo` - "What does this YouTube video say?" → `yt-dlp` extracts subtitles - "Search Bilibili for AI tutorials" → `bili search` (no login required) - "Search the entire web for LLM framework comparisons" → Exa semantic search - "Subscribe to this RSS feed" → `feedparser` parses **No need to remember commands.** After the Agent reads SKILL.md, it knows which tool to use. For platforms that require login (Xiaohongshu, Twitter, Reddit, Facebook, Instagram), tell the Agent "Help me configure XXX" to unlock. --- ## Capability boundaries: reading content vs. operating web pages Some tasks go beyond "reading" - logging in, web page operations, form submissions, multi-account isolation, parallel browser sessions, automated login, verification, and high-friction manual intervention. For such "hands-on" scenarios, consider using browser automation tools like [BrowserAct](https://www.browseract.ai/Agent) - 30+ pre-built platform skills, supporting Claude Code / OpenClaw / Cursor and other mainstream Agents. --- ## Design Philosophy **Agent Reach is a capability layer, not just another tool.** It's one layer above specific implementations - responsible for **selection, installation, testing, and routing**, not the underlying reading itself. Reading is done by the Agent directly calling upstream tools, without a packaging layer. When setting up a new Agent environment, you spend time finding tools, installing dependencies, and configuring settings - what Twitter reader to use? How to log in to Reddit? Xiaohongshu's CLI is discontinued, what to replace it with? Agent Reach does one thing: **the most stable connection method, selected, installed, and tested for you. You don't need to worry about upgrades.** ### 🔌 Each platform = primary + backup ordered list of backends Changing connection methods = adjusting the list order, not rewriting code. `agent-reach doctor` tells you which backend is currently in use for each platform. ``` channels/ ├── web.py → Jina Reader ├── twitter.py → twitter-cli ▸ OpenCLI ▸ bird ├── youtube.py → yt-dlp ├── github.py → gh CLI ├── bilibili.py → bili-cli ▸ OpenCLI ▸ search API (yt-dlp was blocked by Bilibili security, retired) ├── reddit.py → OpenCLI ▸ rdt-cli (no zero-configuration path, must be logged in) ├── facebook.py → OpenCLI (desktop browser login state) ├── instagram.py → OpenCLI (desktop browser login state) ├── xiaohongshu.py → OpenCLI ▸ xiaohongshu-mcp ▸ xhs-cli ├── linkedin.py → linkedin-mcp ▸ Jina Reader ├── rss.py → feedparser ├── exa_search.py → Exa via mcporter └── __init__.py → channel registration (for doctor detection) ``` Each channel file **probes each candidate backend in order** (not just checking if the command exists); the first fully functional one is selected; failed ones provide repair prescriptions. Actual reading and searching are done by the Agent directly calling upstream tools. ### Current selection | Scenario | Primary | Backup | Why this selection | |------|------|------|-----------| | Read web pages | [Jina Reader](https://github.com/jina-ai/reader) | — | Free, no API key required | | Read Twitter | [twitter-cli](https://github.com/public-clis/twitter-cli) | [OpenCLI](https://github.com/jackwener/opencli) | Tested for stable search; OpenCLI uses browser login state as a fallback | | Reddit | [OpenCLI](https://github.com/jackwener/opencli) (desktop) | [rdt-cli](https://github.com/public-clis/rdt-cli) | Anonymous interface blocked, official API approval required - only login state routes remain | | Facebook | [OpenCLI](https://github.com/jackwener/opencli) (desktop) | — | Graph API/Groups API permissions tightened; browser login state is the most practical path | | Instagram | [OpenCLI](https://github.com/jackwener/opencli) (desktop) | Official Graph API (Business/Creator + approval) | instaloader-like paths unstable; OpenCLI reuses real browser sessions | | YouTube subtitles + search | [yt-dlp](https://github.com/yt-dlp/yt-dlp) | — | 154K stars, YouTube remains the best (note: no longer used for Bilibili) | | Bilibili | [bili-cli](https://github.com/public-clis/bilibili-cli) | OpenCLI ▸ search API | yt-dlp blocked by Bilibili security (412 in June 2026), bili-cli searchable and readable without login | | Full-text search | [Exa](https://exa.ai) via [mcporter](https://github.com/nicobailon/mcporter) | — | AI semantic search, MCP access without API key | | GitHub | [gh CLI](https://cli.github.com) | — | Official tool, complete API capabilities after authentication | | Read RSS | [feedparser](https://github.com/kurtmckee/feedparser) | — | Python ecosystem standard choice | | Xiashu | [OpenCLI](https://github.com/jackwener/opencli) (desktop) | [xiaohongshu-mcp](https://github.com/xpzouying/xiaohongshu-mcp) (server) ▸ xhs-cli | xhs-cli author switched to OpenCLI (24K stars); browser login state zero-friction | | LinkedIn | [linkedin-scraper-mcp](https://github.com/stickerdaniel/linkedin-mcp-server) | Jina Reader | MCP service, browser automation | > 📌 These are the **current selections**, based on real-machine testing and regular review. If a path fails, we switch to the next one - `agent-reach doctor` always tells you which one is currently in use. --- ## Security Agent Reach prioritizes security in its design: | Measures | Description | |------|------| | 🔒 **Credential local storage** | Cookies and tokens stored locally in `~/.agent-reach/config.yaml`, file permissions 600 (owner-only readable and writable), not uploaded or shared | | 🛡️ **Safe mode** | `agent-reach install --safe` does not automatically modify, only lists what is needed, and lets you decide what to install | | 👀 **Completely open-source** | Transparent code, reviewable at any time. All dependency tools are also open-source projects | | 🔍 **Dry Run** | `agent-reach install --dry-run` previews all operations without making changes | | 🧩 **Pluggable architecture** | If you don't trust a component, replace the corresponding channel file without affecting others | ### 🍪 Cookie Security Suggestions > ⚠️ **Account Risk Warning:** Platforms that use cookies for login (Twitter, Xiaohongshu, etc.), through script/API calls **exist the risk of being detected and banned by the platform**. Please use a **dedicated account**, not your main account. For platforms that require cookies or login status (Twitter, Xiaohongshu, Reddit, Facebook, Instagram, etc.), it is recommended to use a **dedicated account**, not your main account. There are two reasons: 1. **Account Risk** — The platform may detect non-normal browser API call behavior, leading to account restrictions or bans. 2. **Security Risk** — Cookies are equivalent to complete login permissions. Using a dedicated account can limit the impact scope in case of credential leakage. ### 📦 Installation Methods | Method | Command | Suitable Scenarios | |------|------|---------| | One-click automatic (default) | `agent-reach install --env=auto` | Personal computer, development environment | | Safe mode | `agent-reach install --env=auto --safe` | Production server, multi-user machine | | Preview only | `agent-reach install --env=auto --dry-run` | Preview what will be done | ### 🗑️ Uninstallation ```bash agent-reach uninstall ``` Will clear: `~/.agent-reach/` (including all tokens/cookies), each Agent's skill files, and MCP configuration in mcporter. ```bash # Preview only, do not actually delete agent-reach uninstall --dry-run # Delete skill files only, keep token configuration (for reinstallation) agent-reach uninstall --keep-config ``` Uninstall Python package itself: `pip uninstall agent-reach` ## ⭐ Why Star I use this project every day, so I will maintain it. - I will add new features or channels as needed - I will ensure each channel is **usable, easy to use, and free** - If platforms change their anti-scraping measures or APIs, I will find solutions Contribute to Web 4.0 infrastructure. Star this project, and you'll find it next time you need it. ⭐ ## Acknowledgements [OpenCLI](https://github.com/jackwener/opencli) · [twitter-cli](https://github.com/public-clis/twitter-cli) · [rdt-cli](https://github.com/public-clis/rdt-cli) · [xiaohongshu-mcp](https://github.com/xpzouying/xiaohongshu-mcp) · [xhs-cli](https://github.com/jackwener/xiaohongshu-cli) · [bili-cli](https://github.com/public-clis/bilibili-cli) · [yt-dlp](https://github.com/yt-dlp/yt-dlp) · [Jina Reader](https://github.com/jina-ai/reader) · [Exa](https://exa.ai) · [mcporter](https://github.com/nicobailon/mcporter) · [feedparser](https://github.com/kurtmckee/feedparser) · [linkedin-scraper-mcp](https://github.com/stickerdaniel/linkedin-mcp-server) ## Contact - 📧 **Email:** pnt01@foxmail.com - 🐦 **Twitter/X:** [@Neo_Reidlab](https://x.com/Neo_Reidlab) ### Organizing AI Agent Practices I'm currently organizing AI Agent / AI workflow practices, issues, and opportunities in real-world scenarios. Agent Reach is one of my open-source attempts: making Agents more convenient for reading the internet, doing research, market radar, and data collection, and thus encountering many interesting scenarios. If you're also interested in or trying similar directions, welcome to add me on WeChat for discussion: - You have encountered specific information acquisition, search, monitoring, analysis, or automation issues in your business - You're trying to transform markets, sales, operations, research, content, or data workflows using Agents - You're working on AI Agents, browser automation, data collection, vertical industry tools, or related products - You have observations, cases, or puzzles about Agent implementation Please note: `AI Agent + your direction of interest`. I'll prioritize responding to friends with clear notes and specific scenarios or issues. If suitable, I'll also invite you to relevant discussion groups. <p align="center"> <img src="docs/wechat-group-qr.jpg" width="280" alt="WeChat QR"> </p> > Report bugs and request features on [GitHub Issues](https://github.com/Panniantong/Agent-Reach/issues) for easier tracking. ## License [MIT](LICENSE) ## Friendly Links [Ark Agent Plan Model Subscription Package](https://dis.chatdesks.cn/chatdesk/hsyqAgent-Reach.html) — Integrated with multiple SOTA-level models from ByteDance, covering text, code, image, video, and more. Latest support for MiniMax-M3, DeepSeek-V4 series, GLM-5.2, Doubao-Seed-2.0 series, Kimi-K2.6, and more. All-in-one modal model and Harness upgrade. [Tencent Cloud OpenClaw](https://www.tencentcloud.com/act/pro/intl-openclaw?referral_code=G76Y819A&lang=en&pg=) — Deploy OpenClaw on Tencent Cloud Lighthouse in seconds. Access Agent Reach through dialogue. [AtomGit Mirror](https://atomgit.com/qq_51337814/Agent-Reach) — Agent Reach's AtomGit synchronization mirror for easier domestic access and cloning. ## Star History [](https://star-history.com/#Panniantong/Agent-Reach&Date)