Your AI agent can fetch a web page. Ask it what people say about your competitor on Reddit or X, though, and it hits a login wall, a blocked IP or a $200-a-month API. Agent Reach sets up a free, working path to each platform, so your agent can read and search the places your customers actually talk.
Go straight to Four Jobs for prompts you can paste. Install is one line you give your agent. Read Before You Install before you connect your own accounts.
Agent Reach isn't a scraper. It's an installer and a router. For each platform it picks the open-source tool that works right now (yt-dlp for YouTube, twitter-cli for X, Jina Reader for web pages), installs it, checks it's healthy, and gives your agent a skill file so it knows what to call. When a platform blocks a tool, the maintainer switches to the next one and you run an update.
The best research is in conversations, not on homepages. From the README, here's what stops your agent getting to it:
You could find, install and debug a tool for each one. Or paste one line and let your agent do it.
Each one is a prompt you paste into your agent once it's installed. Swap in your own competitor, niche or problem.
Testimonials on their site are picked by them. Reddit threads and X replies aren't. Turn them into a ranked list of complaints you can build against.
Search Reddit and X for posts about [Competitor] from the last 3 months. Group the complaints by theme, rank them by how often they come up, and quote the 3 sharpest with links.
Your best landing page line is probably something a customer already wrote. Get your agent to collect them word for word, not tidied up.
Find 20 Reddit threads where someone asks for help with [problem]. Pull the exact phrases they use to describe it, what they've already tried, and why it didn't work. Quote, don't paraphrase.
Pull the transcripts of the top videos in your niche and the recent posts of accounts you rate. Study the hooks and angles instead of guessing.
Search YouTube for the top 5 videos on [topic] and get their subtitles. For each one, list the hook it opens with and the main points it covers. Then tell me what none of them cover.
Search X, Reddit and GitHub, plus the RSS feeds you follow, in one pass. Good for content ideas, and for spotting a gap before it gets crowded.
Search X, Reddit and GitHub for posts and repos about [niche] from the past 7 days. What are people excited about, what are they complaining about, and what's getting traction?
Always ask for links. Then you can check a quote yourself before it ends up in your pitch deck or on your homepage.
Zero config works straight after install. The rest need you to log in once, and only if you want that platform.
gh for issues and PRs.Zero ConfigPlus five Chinese platforms (Bilibili, XiaoHongShu, V2EX, Xueqiu and the Xiaoyuzhou podcast app), which is where the project started. Worth a look if you sell into China. "Desktop Login" means it borrows the Chrome session you're already logged into.
Free on GitHub at Panniantong/Agent-Reach. Works with Claude Code, Cursor, Codex, Windsurf or any agent that can run commands. You need Python 3.10 or later.
Install Agent Reach: https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md
It installs the package, checks your machine and tells you what's ready. It won't touch system packages unless you approve --system.
agent-reach doctor
Lists every platform as ready or not, and how to fix the ones that aren't.
help me set up Reddit
Say it to your agent and it walks you through the login. Same for X, LinkedIn and the rest. Takes about 2 minutes each.
npx skills add Panniantong/Agent-Reach@agent-reach
The skill installs the CLI the first time it needs it.
pip install https://github.com/Panniantong/agent-reach/archive/main.zip
agent-reach install --env=auto
Update Agent Reach: https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/update.md
Reddit, X, Instagram and Facebook run as you. Platforms don't like automated use and can restrict accounts that act like bots. Keep it to research-sized volumes, or use a second account. Your cookies stay on your machine.
Platforms block unofficial tools all the time. Switching to whatever still works is the whole point of the project, but there can be a gap. When something fails, run agent-reach doctor, then update.
The README is blunt about it: no coin, no token, no wallet, no fee claims. Anything using the name to sell you one is a scam.
No API fees and an MIT license. Web pages go through Jina Reader and search goes through Exa, so those requests leave your machine. You only need a $1-a-month proxy if your network blocks Reddit or X.