Agent Reach: What the Open-Source AI Scraper Really Needs to Run
By Matt Cullerton, founder of Mavric and Apso
Agent Reach is a free, MIT-licensed Python project by GitHub user Panniantong that connects an AI agent to web pages, YouTube, GitHub, Twitter, Reddit, LinkedIn and other platforms through one command-line tool. A TikTok from @johann_andre says it lets an agent scrape "the entire internet for free without any API keys" and that it has "close to 100,000 stars." The project's own docs say six channels work with zero setup, most of the rest need a logged-in session or a cookie, and the star count was 81,624 on the day the video posted.
What is Agent Reach?
The repository is Panniantong/Agent-Reach, created on 2026-02-24. Its description reads "Give your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu, one CLI, zero API fees."
It picks a backend for each platform and checks that it works. Web pages go through Jina Reader, YouTube through yt-dlp, GitHub through the gh CLI, and Twitter through twitter-cli, with OpenCLI and bird as fallbacks. The command agent-reach doctor reports which backend each platform is using.
The main README is in Chinese. The page shown in the video is the English translation at docs/README_en.md.
The video's creator, Johann Sathianathen, did not build it. The contributor list is headed by Panniantong with 325 commits, and no contributor resembles his name. The video never shows the repository URL or the author's handle.
Is Agent Reach free?
Yes. There is no paid tier, license fee or vendor account. The README's cost note says the only possible expense is a server proxy at about one dollar a month, needed only on restricted networks and not on a local machine.
Free and keyless are different things, though, and the README itself mixes them up in one place. Two tables say the Exa web-search channel needs "no API key," while the sample doctor output in the same file says "sign up at exa.ai for free key." I did not install or run the tool, so I cannot say which statement is current.
Does it really need no API keys or logins?
Only for some platforms. The English README's Supported Platforms table marks Web, YouTube, RSS, V2EX, Bilibili and GitHub as "Zero config." Everything else carries a setup requirement:
- Twitter/X is marked "Cookie." The README says direct
twittercommands "still requireTWITTER_AUTH_TOKENandTWITTER_CT0," session tokens copied from a logged-in browser that grant full account access. - Reddit is marked "No zero-config path (anonymous endpoints blocked)."
- Facebook and Instagram are "Desktop only" and reuse "your logged-in Chrome session."
- XiaoHongShu needs OpenCLI or an MCP server.
- Xiaoyuzhou Podcast is marked "Free API key," for transcription through Groq Whisper.
The TikTok's written description is more accurate than its voiceover. It says the tool can "read web pages, pull YouTube transcripts & work with GitHub," which are three of the six zero-config channels.
The keyless web path is genuine. Jina Reader's README says "Feel free to use Reader API in production. It is free, stable and scalable." It also says "Anonymous traffic is the most aggressively rate-limited and lands in the lowest-trust pool," so keyless use is throttled.
Can Agent Reach scrape LinkedIn?
Only public pages, unless you log in. The video stresses "even LinkedIn," but out of the box the LinkedIn channel falls back to Jina Reader on public pages. Full profiles, company pages and job search go through linkedin-mcp-server, which reuses "a LinkedIn session from a signed-in local browser" and warns that LinkedIn "may require a login confirmation in the LinkedIn mobile app" and "may show a captcha challenge during login."
Can using it get your account banned?
The project says yes. At 15 seconds the video claims "anything you want your agent to do is no longer restricted." The install guide recommends "a dedicated/secondary account rather than your main account" for cookie-based platforms. It names two risks: "platforms may detect non-browser API calls and restrict or ban the account," and "cookies grant full account access."
The LinkedIn backend is blunter. Its FAQ says "LinkedIn's User Agreement prohibits automated access, and accounts using automated tools can be restricted or banned. Use at your own risk; there is no guarantee of account safety." The video's headline use case, generating leads from LinkedIn, is the one the upstream maintainer flags as a terms violation.
How many GitHub stars does Agent Reach have?
An Internet Archive snapshot from 2026-09-15, about nine hours before the video posted, shows 81,624 stars. The next day's snapshot shows 82,244. Earlier snapshots show 66,899 on 2026-08-05 and 75,001 on 2026-08-25. A GitHub API call on 2026-10-19 returned 99,678 stars and 8,805 forks, so "close to 100,000" became accurate about a month after the video.
Trendshift lists the repository as trending on 13 days between 2026-06-05 and 2026-10-05, with its #1 Repository of the Day placement on 2026-10-02.
What the video shows
Every frame shows either the rendered README in a mobile browser or the creator talking. There is no terminal, no doctor run and no scraped output. The channel list on screen matches the README, including the reddit.py line that reads "no zero-config path, login required," cut off at the screen edge.
The video ends by asking viewers to comment "repos" to get the link by direct message. The repository is public and easy to find by name. The creator's LinkedIn headline lists him as co-founder of Cyndra.ai, and he has no stake in Agent Reach itself.
Bottom line
Agent Reach is real, free and useful for reading web pages, YouTube and GitHub without setup. Twitter, Reddit, Facebook, Instagram and full LinkedIn access need logged-in sessions or cookies, and the project's own docs warn that those can get accounts banned. Use a secondary account if you try them.
Sources
- Panniantong/Agent-Reach on GitHub - the repository, license, creation date and contributor list.
- Agent Reach English README - supported platforms table, backend list, cost note and Twitter token requirement.
- Agent Reach install guide - the secondary-account recommendation and ban warning.
- stickerdaniel/linkedin-mcp-server - the LinkedIn backend, its login flow and its terms warning.
- jina-ai/reader - the keyless web-reading backend and its rate limits.
- Internet Archive capture, 2026-09-15 - 81,624 stars on the day the video posted.
- Archive snapshot, 2026-09-16 - 82,244 stars and 7,169 forks.
- Wayback Machine capture, 2026-08-25 - earlier star counts for the growth curve.
- Trendshift repository 24387 - trending history and #1 date.
- Johann Sathianathen on LinkedIn - the creator's name and Cyndra.ai headline.
Published September 15, 2026. Based on a TikTok by @johann_andre. Full video notes (markdown)