What it does

Agent Reach is an installer and diagnostic wrapper for a curated set of free web-scraping tools. You paste one command into Claude Code, Cursor, or another CLI-capable agent, and it installs yt-dlp, twitter-cli, rdt-cli, Jina Reader, and a dozen other upstream utilities. After installation, the agent can fetch YouTube transcripts, read Reddit threads, search GitHub, parse RSS feeds, and browse Chinese platforms like Bilibili and XiaoHongShu. A built-in doctor command checks which channels are working and which need cookies or proxies.

The interesting bit

The project treats itself as disposable scaffolding, not a framework. Each platform is a thin Python file that merely detects whether an upstream tool is installed; the agent calls that tool directly. If Twitter’s CLI breaks tomorrow, you swap the backend without touching Agent Reach’s core. The README is unusually honest about this: “不满意？换掉就行” — don’t like it? Replace it.

Key highlights

• Covers 15+ platforms including Twitter/X, Reddit, YouTube, Bilibili, XiaoHongShu, Douyin, WeChat public accounts, Weibo, V2EX, and Xueqiu
• No API keys for basic usage; semantic web search via Exa through an MCP server
• Cookie-based authentication stays local; the project emphasizes it does not upload credentials
• agent-reach doctor gives a green/red status board for every channel
• OpenClaw users must manually enable exec permissions first — the README flags this explicitly

Caveats

• Several platforms (Twitter search, XiaoHongShu, Douyin, LinkedIn profiles) require cookie export from a logged-in browser session
• Server deployments need a proxy (~$1/month) for Bilibili and some Chinese services
• The project is glue and curation: all heavy lifting is done by upstream tools like yt-dlp (154K stars) and twitter-cli (2.1K stars)

Verdict

Worth a look if you’re tired of manually provisioning scrapers for every new AI agent project. Skip it if you already maintain your own toolchain or need guaranteed uptime — free scrapers break when platforms change, and the README admits it will chase those fixes but can’t promise speed.