Panniantong/Agent-Reach

Agent-Reach is a scaffolding toolkit that gives any AI agent full read-and-search access to the internet — Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu, LinkedIn, WeChat, Weibo, V2EX, and more — through a single CLI install with zero API fees. The core problem it solves is brutal: every platform has its own barriers. Twitter charges per-read API calls, Reddit blocks server IPs with 403s, XiaoHongShu requires login, Bilibili blocks overseas connections, and YouTube subtitles need dedicated tooling. Developers building AI agents spend hours wiring up each platform individually, debugging auth flows, and handling anti-scraping measures. Agent-Reach eliminates all of that with one command. You paste a single install URL into Claude Code, Cursor, Windsurf, or any agent that can run shell commands, and it auto-installs every upstream tool — bird for Twitter, yt-dlp for YouTube and Bilibili, Jina Reader for arbitrary web pages, gh CLI for GitHub, feedparser for RSS, Exa for semantic web search, and MCP servers for XiaoHongShu, Douyin, and LinkedIn. Each channel is a pluggable Python module. Don't trust a particular scraper? Swap it out by replacing one file. The architecture is deliberately thin — Agent-Reach doesn't wrap the upstream tools at runtime. It installs them, registers a SKILL.md so your agent knows which tool handles which platform, then gets out of the way. A built-in doctor command diagnoses which channels are working and which need configuration. Cookie-based auth for Twitter and XiaoHongShu keeps everything free — no paid API keys required. The project ships with safe mode for production servers, dry-run previews, and a clean uninstall command. At 10.5K stars in under a month, it hit a nerve with developers who are tired of reinventing internet access for every new agent project.