# Agent-Reach

> Setup layer that lets AI agents read the web, YouTube, GitHub, Reddit and more

- Repository: https://github.com/Panniantong/Agent-Reach
- Category: Browser & computer use
- Stars: 101,057 · Forks: 8,907 · Language: Python · License: MIT
- Experience: Some experience
- Page: https://ir.adyog.com/repos/panniantong/agent-reach/

AI coding agents can edit code, but they often fail when asked to read the web. YouTube transcripts, Twitter search, Reddit threads and Xiaohongshu posts each need a different tool, login or workaround. Setting all of that up by hand takes a while.

Agent Reach is a capability layer, not a scraper of its own. For each platform it keeps an ordered list of backends, such as twitter-cli first and OpenCLI second. It tests them, picks the first one that really works, and installs it. The agent then calls those upstream tools directly, guided by a SKILL.md file that the project registers.

If a platform blocks a tool, the project moves to the next backend, so users don't have to change anything. The README gives one example: yt-dlp was blocked by Bilibili and was replaced with bili-cli. The `agent-reach doctor` command reports which channels work and which backend each one uses.

## What it does

- **Ordered backends per platform**: Each channel has a preferred tool and fallbacks. The project probes them for real, not just checks that a command exists, and picks the first that works.
- **Zero-config channels**: Web pages (Jina Reader), YouTube subtitles and search (yt-dlp), RSS (feedparser), GitHub (gh CLI), Bilibili search and V2EX work without setup. Exa web search connects over MCP without a key.
- **Login-based platforms**: Twitter/X, Reddit, Facebook, Instagram, Xiaohongshu, LinkedIn and others unlock after setup. You tell the agent to configure them and it guides you through it.
- **Doctor command**: `agent-reach doctor` shows the state of each channel, which backend is active, and how to fix anything that is broken.
- **Safe install defaults**: `agent-reach install` only checks the environment by default. Changing the system needs `--system`, and `--dry-run` previews every action. Cookies stay in a local config file.

## What you could build

- Ask your agent to summarize a YouTube tutorial from its subtitles.
- Search Twitter/X or Reddit for what people say about a product or a bug.
- Have an agent read any web page as clean text instead of raw HTML.
- Search Bilibili or Xiaohongshu and summarize the results.
- Subscribe an agent to RSS feeds and report new items.

## Who it's for

Developers who use agents such as Claude Code, Cursor or OpenClaw and want them to read social and video sites; know that login-based platforms need cookies and a spare account is advised.

## Get started

Paste the install prompt into your agent, which then reads the install guide and sets things up. Afterward, run `agent-reach doctor` to see which channels work. For a safe preview first, use `agent-reach install --env=auto --dry-run`.

```
agent-reach install --env=auto --dry-run
```
```
agent-reach install --env=auto --system
```
```
agent-reach doctor
```

Tags: ai-agents, web-scraping, cli, mcp, youtube