Agent Reach

Agent Reach is an open-source Python CLI tool that equips AI agents with internet-reading capabilities across major platforms — Twitter/X, Reddit, YouTube, GitHub, Bilibili, Xiaohongshu, LinkedIn, and more — without requiring paid APIs. Built and maintained by Panniantong, the project is licensed under MIT and reached over 83,000 GitHub stars, appearing as a Trendshift #1 Repository of the Day.

What It Is

Agent Reach is a capability layer — not a scraper or a wrapper — that sits above individual platform tools and handles tool selection, installation, health checking, and routing for AI agents. When an agent needs to read a tweet, extract a YouTube transcript, or search Reddit, Agent Reach selects the best available backend, installs it, and routes the request. The agent calls upstream tools directly; Agent Reach manages which tool is active and healthy at any given time.

How the Multi-Backend Routing Works

Each supported platform is represented as an ordered list of backends (primary + fallbacks). If a backend breaks — due to platform anti-scraping changes, API deprecations, or authentication shifts — Agent Reach switches to the next candidate without requiring user action. The agent-reach doctor command shows which backend is currently active for each channel and provides repair prescriptions for broken ones.

Example channel routing:

• Twitter/X: twitter-cli → OpenCLI → bird
• Bilibili: bili-cli → OpenCLI → search API (yt-dlp retired after B站 412 blocks in 2026-06)
• Reddit: OpenCLI (desktop) → rdt-cli (no anonymous path; login state required)
• Xiaohongshu: OpenCLI → xiaohongshu-mcp → xhs-cli
• Web: Jina Reader (free, no API key)
• Search: Exa via mcporter (MCP, free, no key)

Setup Path

Installation is designed to be delegated entirely to the AI agent itself. Users copy a single instruction to their agent (Claude Code, OpenClaw, Cursor, Windsurf, or any shell-capable agent):

帮我安装 Agent Reach：https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md

The agent fetches the install manifest, installs the agent-reach CLI from the repository, checks for Node.js, gh CLI, and mcporter, and registers a SKILL.md file so the agent knows which commands to call for each platform. Six zero-configuration channels activate immediately; login-required platforms (Twitter, Reddit, Xiaohongshu, Facebook, Instagram) are unlocked on demand by telling the agent to configure them.

Security Model

Agent Reach stores credentials (cookies, tokens) locally in ~/.agent-reach/config.yaml with 600 file permissions. The install command defaults to read-only environment checking; system modifications only occur when --system is explicitly passed. A --dry-run flag previews all planned operations. The project explicitly warns users to use dedicated secondary accounts for cookie-based platforms (Twitter, Xiaohongshu, Reddit) due to platform ban risk.

Update: v1.5.0 — Multi-Backend Routing + Real Health Checks + OpenCLI

The latest release, v1.5.0 (published June 2026), introduced the capability-layer architecture with ordered multi-backend routing, real backend probing (not just command existence checks), and OpenCLI integration as a browser-session fallback for Reddit, Facebook, Instagram, and Xiaohongshu. The Bilibili channel was updated in this cycle to retire yt-dlp (blocked by B站 412 anti-scraping) and promote bili-cli as the primary backend. The project supports Python 3.10+ and is available in Chinese, English, Japanese, and Korean documentation.