WHAT IS IT?
Agent Reach is a CLI scaffolding that plugs your AI agent into the public web — Twitter/X, Reddit, YouTube, Bilibili, GitHub, Xiaohongshu, Douyin, LinkedIn, WeChat, Weibo, RSS — without forcing you to pay for a single official API. The project doesn't reinvent scraping: it orchestrates proven tools from the ecosystem (yt-dlp, Jina Reader, twitter-cli, rdt-cli, gh, feedparser, Exa…) behind a unified, agent-friendly layer.
Installation is a single command (agent-reach install), which detects your environment and pulls the system dependencies. Coding assistants (Claude Code, Cursor, Codex…) then call it like any other CLI, or consume it over MCP via mcporter.
WHY IS IT INTERESTING?
- Zero paid APIs: No $100/month Twitter key, no YouTube Data API quota. The underlying tools talk straight to public sites, for free.
- Asian platform coverage: Bilibili, Xiaohongshu, Douyin, WeChat, Weibo. Rare in Western projects, invaluable for product or competitive intelligence in China.
- One surface for the agent: Your LLM doesn't have to learn ten different SDKs. It learns one verb (
read,search,watch) and a list of platforms. - Autonomous install: Hand the install URL to your agent and it figures out the binaries (yt-dlp, gh, twitter-cli…) and their config on its own.
- Open Python stack: MIT license, 100% Python, easy to fork or extend to bolt on a new platform.
- Built-in MCP bridge: With
mcporter, the CLI exposes itself as a Model Context Protocol server, consumable by any MCP client.
USE CASES
- Multi-platform product and competitive monitoring driven by an agent
- YouTube or Bilibili transcript extraction to feed a RAG pipeline
- Tracking Reddit threads, tweets or LinkedIn posts on a given topic
- Semantic web search via Exa, embedded inside a research agent
- Collecting RSS and WeChat content for automated daily summaries