Agent Reach

Agent Reach is an internet capability router designed for AI Agents to research, search, and fetch content from 15 platforms.

When to use:

• When a user wants to research a topic, perform a deep dive, or search the web.
• When a user mentions specific platforms: Twitter/X, Reddit, Facebook, Instagram, YouTube, GitHub, Bilibili, XiaoHongShu, Xiaoyuzhou Podcast, LinkedIn, V2EX, Xueqiu, or RSS.

When NOT to use:

• For writing reports, analysis, or translation (this skill is for fetching content only).
• For write operations like posting, commenting, or liking.
• If the user already has a dedicated skill installed for a specific platform (prefer the dedicated skill).

Operational Rules:

• Health-check first: For login-backed platforms (XiaoHongShu, Reddit, Bilibili, Twitter, Facebook, Instagram), always run agent-reach doctor --json to identify the active_backend before attempting commands.
• Announce usage: Explicitly state which platform and backend you are using (e.g., "using agent-reach, platform X via backend Y").
• Workspace hygiene: Never create files in the agent workspace. Use /tmp/ for temporary files and ~/.agent-reach/ for persistent data.

Agent Reach acts as a capability layer that manages the selection, installation, and routing of backends (CLIs, APIs, and MCP tools) for various platforms. It does not wrap the tools; instead, it ensures the most reliable access path is available so your Agent can call the upstream tools directly.

Each platform (channel) maintains an ordered list of primary and fallback backends. When a backend becomes blocked or deprecated (e.g., yt-dlp being blocked on Bilibili), Agent Reach automatically re-routes to the next available working backend in the list.

Key Platform Routing Examples:

• Twitter: twitter-cli $\rightarrow$ OpenCLI $\rightarrow$ bird
• Bilibili: bili-cli $\rightarrow$ OpenCLI $\rightarrow$ search API (Note: yt-dlp is retired due to 412-blocking)
• Reddit: OpenCLI $\rightarrow$ rdt-cli (Requires login)
• XiaoHongShu: OpenCLI $\rightarrow$ xiaohongshu-mcp $\rightarrow$ xhs-cli
• LinkedIn: linkedin-mcp $\rightarrow$ Jina Reader
• Web Pages: Jina Reader
• YouTube: yt-dlp
• GitHub: gh CLI
• Web Search: Exa via mcporter

Agent Reach is an internet capability router used when a user wants to research, search, or look up information online. It provides access to 15 platforms through multiple backends (CLIs, APIs, MCP tools).

When to use

• Research/Search tasks: e.g., "research this topic", "search for X", "what are people saying about X".
• Platform-specific access: When a user mentions or shares a URL for Xiaohongshu (XHS), Twitter/X, Bilibili, Reddit, Facebook, Instagram, V2EX, LinkedIn, YouTube, GitHub, Xiaoyuzhou podcasts, Xueqiu, or RSS feeds.

Constraints

• Do NOT use for: Content processing (writing reports, data analysis, translation), write operations (posting, commenting, liking), or platforms that already have a dedicated specialized skill.
• Workspace Rule: Do not create files in the agent workspace. Use /tmp/ for temporary output and ~/.agent-reach/ for persistent data.

Agent Reach provides access to various platforms using specialized backends. Key supported scenarios include:

• Web Content: Uses Jina Reader for free web page reading without API keys.
• Twitter/X: Uses twitter-cli (cookie-based) for searching, reading tweets, timelines, and profiles without API costs.
• Reddit: Uses rdt-cli (cookie-based) for searching and reading full posts and comments.
• Video & Subtitles: Uses yt-dlp for YouTube, Bilibili, and 1800+ other sites to extract metadata and subtitles.
• Social Media:
  • XiaoHongShu (XHS): Uses xhs-cli for searching, reading notes, and comments.
  • Douyin: Uses douyin-mcp-server for parsing video info and downloading without watermarks.
  • Weibo: Accessible via mcporter commands.
• Search: Uses Exa via mcporter for AI semantic search.
• Other: gh CLI for GitHub, feedparser for RSS, and linkedin-scraper-mcp for LinkedIn automation.

Agent Reach prioritizes security through local credential storage and transparent architecture:

• Local Storage: Cookies and tokens are stored locally in ~/.agent-reach/config.yaml with file permissions set to 600 (read/write only by the owner). They are never uploaded to a central server.
• Safe Installation: The --safe flag allows you to see what needs to be installed without making automatic system changes.
• Dry Run: The --dry-run flag allows you to preview all installation operations without making any changes.

⚠️ Account Safety Warning

When using platforms that require Cookie-based login (such as Twitter, Xiaohongshu, Reddit, Facebook, or Instagram), there is a risk that the platform may detect non-browser API activity and ban the account.

Recommendation: Always use a dedicated secondary account (burner account) for Agent Reach rather than your primary personal account to mitigate the risk of account suspension or credential exposure.

Install the core Agent Reach infrastructure, which includes tools like gh CLI, Node.js, mcporter, Exa search, and yt-dlp configuration. This enables zero-config channels for Web (Jina Reader), YouTube, GitHub, RSS, Exa Search, V2EX, and basic Bilibili.

Recommended: Using pipx

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

Using a Virtual Environment (for Homebrew/PEP 668 environments)

python3 -m venv ~/.agent-reach-venv
source ~/.agent-reach-venv/bin/activate
pip install https://github.com/Panniantong/agent-reach/archive/main.zip
agent-reach install --env=auto

Windows (PowerShell)

If python3 points to the Microsoft Store, use the Python Launcher py -3:

py -3 -m venv $env:USERPROFILE\.agent-reach-venv
$env:USERPROFILE\.agent-reach-venv\Scripts\Activate.ps1
python -m pip install https://github.com/Panniantong/agent-reach/archive/main.zip
agent-reach install --env=auto