AI analysis grounded in the code graph — computed facts, not vibes · 2026-09-23T03:03:03Z
Agent Reach is a Python CLI (and MCP server) that gives an AI agent read/search access to a dozen-odd web platforms — Twitter/X, Reddit, YouTube, GitHub, Bilibili, Xiaohongshu, LinkedIn, V2EX, Xueqiu, RSS and plain web pages — without requiring paid platform APIs. It exposes per-platform check()/read() functions in agent_reach/channels/* and a diagnostics command (agent_reach doctor) that reports which channels are working, aimed at developers wiring agent frameworks (Claude Code, Cursor, Windsurf, OpenClaw) into external data sources.
The graph facts don't fully explain a 659-star single-day jump: the repo was only created 2026-02-24, and the visible signal is a fast release cadence (v1.3.0 → v1.5.0 in about ten weeks) plus README claims of zero-config, one-line install/update and "we fix broken channels so you don't have to." These are plausible growth drivers but are self-reported claims rather than confirmed adoption metrics, so the correlation should be read cautiously.
What changed recently, how it's actually built (from the code graph), and whether you should care. Free account — no card, no spam.