agent-reach

name	agent-reach
description	Internet access toolkit for AI agents. Read & search Twitter/X, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu, LinkedIn, Boss, RSS, and any web page. Use when the user asks to read a tweet, search Twitter, watch/summarize a YouTube or Bilibili video, search Reddit, read XiaoHongShu posts, browse the web, search the internet, read RSS feeds, or access any online content. Triggers: 'read this tweet', 'search Twitter', 'summarize this video', 'search Reddit', 'read this web page', 'search the internet', 'read XiaoHongShu', 'check this GitHub repo', 'read RSS', 'search YouTube', 'search Bilibili'.
homepage	https://github.com/Panniantong/Agent-Reach
metadata	{"openclaw":{"emoji":"🌐","requires":{"anyBins":["curl","bird","yt-dlp","gh","mcporter"]},"install":[{"id":"uv","kind":"uv","package":"https://github.com/Panniantong/agent-reach/archive/main.zip","bins":["agent-reach"],"label":"Install agent-reach (pip)"}]}}

Agent Reach - Internet Access Toolkit

Give your AI agent access to the entire internet. This skill enables reading and searching across major platforms with zero API fees.

Supported Platforms & Commands

Web Pages (always available)

Read any web page using Jina Reader:

curl -s "https://r.jina.ai/URL"

• Works on any public URL
• Returns clean markdown, not raw HTML
• No API key needed

Twitter / X

Requires: bird CLI + Twitter cookies configured.

# Read a single tweet
bird read "https://x.com/user/status/123" --json

# Search tweets
bird search "keyword" --json

# View user timeline
bird timeline @username --json

Cookie setup: Use Cookie-Editor Chrome extension on x.com, export as Header String, then:

agent-reach configure twitter-cookies "PASTED_COOKIE_STRING"

YouTube

Requires: yt-dlp

# Get video metadata + subtitles
yt-dlp --dump-json "https://youtube.com/watch?v=xxx"

# Extract subtitles only
yt-dlp --write-sub --write-auto-sub --sub-lang en --skip-download -o "%(title)s" "URL"

# Search YouTube
yt-dlp "ytsearch10:keyword" --dump-json --flat-playlist

Bilibili

Requires: yt-dlp (same tool as YouTube)

# Get video info + subtitles
yt-dlp --dump-json "https://www.bilibili.com/video/BVxxx"

# Search Bilibili
yt-dlp "bilisearch10:keyword" --dump-json --flat-playlist

Note: On servers, may need a proxy configured via agent-reach configure proxy URL.

GitHub

Requires: gh CLI

# View a repository
gh repo view owner/repo

# Search repositories
gh search repos "LLM framework" --limit 10

# Search code
gh search code "function_name" --repo owner/repo

# List issues
gh issue list --repo owner/repo

# Read an issue
gh issue view 123 --repo owner/repo

For private repos: gh auth login

Reddit

# Read a post + comments (JSON API)
curl -s -H "User-Agent: AgentReach/1.0" "https://www.reddit.com/r/subreddit/comments/postid.json"

# Search via Exa (semantic search, better quality)
mcporter call 'exa.web_search_exa(query: "site:reddit.com your search query", numResults: 10)'

On servers: may need proxy for direct Reddit access.

Internet Search (Exa)

Requires: mcporter with Exa configured.

# Semantic web search
mcporter call 'exa.web_search_exa(query: "your search query", numResults: 10)'

# Search with content extraction
mcporter call 'exa.web_search_exa(query: "your query", numResults: 5, textContentsOptions: "text")'

XiaoHongShu (via MCP)

Requires: Docker + xiaohongshu-mcp running.

# Search posts
mcporter call 'xiaohongshu.search_feeds(keyword: "keyword", page: 1)'

# Read a specific post
mcporter call 'xiaohongshu.get_feed_detail(note_id: "NOTE_ID")'

# Get comments
mcporter call 'xiaohongshu.get_feed_comments(note_id: "NOTE_ID")'

Setup:

docker run -d --name xiaohongshu-mcp -p 18060:18060 xpzouying/xiaohongshu-mcp
mcporter config add xiaohongshu http://localhost:18060/mcp

Then open http://localhost:18060 to scan QR code for login.

LinkedIn (via MCP)

Requires: linkedin-scraper-mcp

# Get person profile
mcporter call 'linkedin.get_person_profile(linkedin_url: "https://linkedin.com/in/username")'

# Search jobs
mcporter call 'linkedin.search_jobs(keywords: "software engineer", location: "San Francisco")'

Basic public pages can also be read via Jina Reader: curl -s "https://r.jina.ai/https://linkedin.com/in/username"

Boss (via MCP)

Requires: mcp-bosszp

# Search jobs
mcporter call 'bosszhipin.search_jobs_tool(keyword: "Python", city: "beijing")'

# Send greeting to HR
mcporter call 'bosszhipin.start_chat_tool(job_id: "JOB_ID", greeting: "Hello")'

RSS Feeds

python3 -c "
import feedparser
d = feedparser.parse('https://example.com/feed.xml')
for e in d.entries[:5]:
    print(f'{e.title}: {e.link}')
"

Diagnostics

Check which channels are available:

agent-reach doctor

Installation

If agent-reach is not installed:

pip install https://github.com/Panniantong/agent-reach/archive/main.zip
agent-reach install --env=auto

Safe mode (no auto system changes):

agent-reach install --env=auto --safe

Important Notes

• All tools are free and open-source. No paid API keys required.
• Cookie-based platforms (Twitter, XiaoHongShu): use a dedicated/secondary account for safety.
• Proxy (~$1/month) only needed on servers for Reddit/Bilibili access.
• Local machines generally work without proxy.
• When a platform returns errors, run agent-reach doctor to diagnose.