Self-description (github.com/Panniantong/Agent-Reach): “Give your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.”
Agent-Reach is an MIT-licensed Python CLI, created on GitHub on February 24, 2026, that lets AI agents read and search platforms such as X, Reddit, YouTube, GitHub, Bilibili and XiaoHongShu without paid API fees. It had 95,839 stars and 8,392 forks by October 11, 2026. Star history shows three surges: about 9,200 stars in its first three weeks, about 20,000 in June 2026 when it first hit GitHub Trending and reached #1, and about 7,700 in early October during a second trending peak. The project installs by handing an AI agent a one-sentence prompt pointing to an install doc. It frames itself as a layer that picks, installs, health-checks and reroutes between free upstream tools.
Two loops appear in the evidence. First, trending lists. The repo first appeared on GitHub Trending on June 5, 2026, reached #1 on June 15, and trended for 32 days. It returned to #1 on October 2, and Trendshift ranked it in its top results on several days. These periods overlap with its largest star surges. Trending exposure likely brought new visitors, and their stars helped keep it ranked, though the claims do not measure this directly. Second, reposting. Reddit posts repeated the zero-API-fee tagline along with growing star counts (23K, then 42.7K, then 88k), and the README asks visitors to star the project. These loops are reconstructed from timing and do not have measured cause and effect.
Moderate inference| Date | Type | Event |
|---|---|---|
| Feb 24, 2026 | Repository | Panniantong/Agent-Reach repository created |
| Feb 25, 2026 | Release | First release v1.1.0 |
| Feb 26, 2026 | Release | Release v1.2.0 |
| March 2026 | Strategy | Routing changed after single-platform CLIs went offline |
| Mar 04, 2026 | Release | Release v1.3.0 |
| Mar 24, 2026 | On Reddit: 🔎 Open Source AI Resource List (curated, ongoing) | |
| Mar 29, 2026 | On Reddit: 🔎 AI Agents Resource List (curated, ongoing) | |
| Mar 29, 2026 | On Reddit: "Give your ai agent eyes to see the entire internet for free - Read & search - Twitter - Reddit - YouTube - GitHub - Bilibili - XiaoHongShu - One CLI, zero API fees." ➡️ People are saying this is a GEM! What do you think? | |
| Mar 29, 2026 | On Reddit: "Give your ai agent eyes to see the entire internet for free - Read & search - Twitter - Reddit - YouTube - GitHub - Bilibili - XiaoHongShu - One CLI, zero API fees." ➡️ Do you think this is useful? People are calling it a GEM! | |
| Apr 06, 2026 | On Reddit: Agent-Reach: This new github repo gives any AI Agent instant access to the entire web without API keys | |
| June 2026 | Reversal | Removed non-working entry points; Bilibili backend switched to bili-cli |
| Jun 05, 2026 | Milestone | First appearance on GitHub Trending (Python #3) |
| Jun 07, 2026 | On Reddit: Local agents on a MacBook Pro M5 finally feel practical to me | |
| Jun 10, 2026 | Release | Release v1.4.1 |
| Jun 10, 2026 | Release | Release v1.4.2 |
| Jun 11, 2026 | Release | Release v1.5.0 |
| Jun 15, 2026 – Jun 18, 2026 | Milestone | Reached #1 on GitHub Trending |
| Jun 16, 2026 | On Reddit: "your agent can search Twitter, Reddit, and GitHub for free - zero API keys, zero billing 😳 agent-reach is trending on github 23K stars. lets your AI agent read Twitter posts, browse Reddit threads, search GitHub repos, watch YouTube videos" ➡️ a channel map for agent web access? | |
| Jun 16, 2026 | On Reddit: "your agent can search Twitter, Reddit, and GitHub for free - zero API keys, zero billing 😳 agent-reach is trending on github 23K stars. lets your AI agent read Twitter posts, browse Reddit threads, search GitHub repos, watch YouTube videos" ➡️ turns agent web access into a channel map? | |
| Jun 23, 2026 | On Reddit: A study scanned 42,000 AI agent skills. A quarter were vulnerable, 5% likely malicious. Here is how to extend your coding agent without getting burned | |
| Jun 29, 2026 | On Reddit: "your agent can search Twitter, Reddit, YouTube, GitHub, B站, 小红书, LinkedIn, and the open web for free - zero API keys, zero billing 😳 agent-reach now supports 14+ platforms with self-healing routing. one of them dies? it switches to the next backend automatically. 42.7K stars" ➡️ you using? | |
| Jun 29, 2026 | On Reddit: "your agent can search Twitter, Reddit, YouTube, GitHub, B站, 小红书, LinkedIn, and the open web for free - zero API keys, zero billing 😳 agent-reach now supports 14+ platforms with self-healing routing. one of them dies? it switches to the next backend automatically. 42.7K stars" ➡️ useful? | |
| Jun 29, 2026 | Press | ToKnow.ai article on Agent-Reach |
| Jun 30, 2026 | On Reddit: Distribution is the whole game, but what about the API costs? is social listening actually in your pipeline? | |
| Jul 24, 2026 | On Reddit: Show Reddit: AI Workforce OS — Open-Source Autonomous AI Organization & Multi-Agent Orchestrator (Ollama, OpenHands, OpenClaw, ChatDev, RTK Token Killer & Real Obsidian Memory) | |
| Aug 02, 2026 | Hacker News | On Hacker News: Anyone used Agent Reach yet? any good? |
| Aug 07, 2026 | Press | Techbytes article on Agent Reach |
| Aug 07, 2026 | Hacker News | On Hacker News: Agent Reach: An open-source CLI that gives AI agents access to the internet |
| Sep 09, 2026 | On Reddit: 7 small agent skills that change how I use coding agents | |
| Sep 10, 2026 | On Reddit: Turn Hermes Into an AI Team, Not One Giant Agent | |
| Sep 13, 2026 | Milestone | Trendshift #6 overall, #2 in Python |
| Sep 16, 2026 | On Reddit: 5 open-source repos to let AI agents collect data from the web | |
| Oct 02, 2026 – Oct 04, 2026 | Milestone | Second trending peak in early October |
| Oct 03, 2026 | On Reddit: Agent-Reach: give your AI coding agent access to Twitter, Reddit, YouTube, GitHub and more (88k stars) | |
| Oct 06, 2026 | Milestone | Ranked #5 on GitHub Trending |
| Oct 08, 2026 | On Reddit: Agent Reach 1.5.0 on a fresh Windows machine: 5 of 16 channels, one green tick that is wrong, and a lot of work delegated to other tools |
+9,239 stars in 21 day(s); peak day Mar 04 with +814. Afterwards (days 31–90) it averaged 189.8 stars/day, which stayed well above the 1.0/day before the episode. 3 public event(s) observed within this window (timing only).
+19,995 stars in 21 day(s); peak day Jun 15 with +1,796. Afterwards (days 31–90) it averaged 412.7 stars/day, which stayed well above the 80.4/day before the episode. 9 public event(s) observed within this window (timing only).
+7,737 stars in 8 day(s); peak day Oct 02 with +1,466. Too recent to say whether the faster pace lasted. 4 public event(s) observed within this window (timing only).
| When | Launch episode | Stars before | +24h | +7 days | +30 days | +90 days |
|---|---|---|---|---|---|---|
| Mar 24, 2026 | Launch activity, Mar 2026 | 10K | +217 | +3.5K | +7.3K | +27K |
| Mar 29, 2026 | Launch activity, Mar 2026 | 12K | +1.7K | +3.1K | +6.1K | +31K |
| Apr 06, 2026 | Launch activity, Apr 2026 | 15K | +970 | +1.9K | +3.5K | +36K |
| Jun 07, 2026 – Jun 11, 2026 | Launch activity, Jun 2026 | 22K | +2K | +6.3K | +30K | +56K |
| Jun 16, 2026 | Launch activity, Jun 2026 | 30K | +2.7K | +7.9K | +26K | +51K |
| Jun 23, 2026 | Launch activity, Jun 2026 | 37K | +2.1K | +9.3K | +22K | +47K |
| Jun 29, 2026 – Jun 30, 2026 | Launch activity, Jun 2026 | 44K | +2.6K | +7.4K | +18K | +42K |
| Jul 24, 2026 | Launch activity, Jul 2026 | 60K | +555 | +3.3K | +14K | — |
| Sep 09, 2026 – Sep 10, 2026 | Launch activity, Sep 2026 | 78K | +461 | +3.7K | +17K | — |
| Sep 16, 2026 | Launch activity, Sep 2026 | 82K | +813 | +3.1K | — | — |
| Oct 03, 2026 | Launch activity, Oct 2026 | 89K | +2.2K | +6.7K | — | — |
| Oct 08, 2026 | Launch activity, Oct 2026 | 94K | +1.6K | — | — | — |
Each row's "state before" and outcome details are in the evidence drawer. Comparing rows shows how the project's situation and response differed between launches.
1 related public event(s): reddit post. State before: About 10,214 stars the day before; repository 28 days old; first launch episode Pigtail found.
Stars went from 10,214 the day before to 37,275 (+24h: +217, +48h: +348, +7d: +3,533, +30d: +7,348, +90d: +27,061). Observed after the launch; not proof that the launch caused the change.
3 related public event(s): reddit post. State before: About 11,791 stars the day before; repository 33 days old; 1 earlier launch episode(s).
Stars went from 11,791 the day before to 42,808 (+24h: +1,706, +48h: +1,956, +7d: +3,059, +30d: +6,115, +90d: +31,017). Observed after the launch; not proof that the launch caused the change.
1 related public event(s): reddit post. State before: About 14,850 stars the day before; repository 41 days old; 2 earlier launch episode(s).
Stars went from 14,850 the day before to 50,655 (+24h: +970, +48h: +1,154, +7d: +1,899, +30d: +3,520, +90d: +35,805). Observed after the launch; not proof that the launch caused the change.
4 related public event(s): reddit post, release. State before: About 22,069 stars the day before; repository 103 days old; 3 earlier launch episode(s).
Stars went from 22,069 the day before to 77,783 (+24h: +2,011, +48h: +3,055, +7d: +6,312, +30d: +29,919, +90d: +55,714). Observed after the launch; not proof that the launch caused the change.
2 related public event(s): reddit post. State before: About 30,177 stars the day before; repository 112 days old; 4 earlier launch episode(s).
Stars went from 30,177 the day before to 81,163 (+24h: +2,705, +48h: +3,658, +7d: +7,875, +30d: +26,276, +90d: +50,986). Observed after the launch; not proof that the launch caused the change.
1 related public event(s): reddit post. State before: About 37,275 stars the day before; repository 119 days old; 5 earlier launch episode(s).
Stars went from 37,275 the day before to 84,162 (+24h: +2,133, +48h: +3,424, +7d: +9,282, +30d: +22,242, +90d: +46,887). Observed after the launch; not proof that the launch caused the change.
3 related public event(s): reddit post. State before: About 43,977 stars the day before; repository 125 days old; 6 earlier launch episode(s).
Stars went from 43,977 the day before to 85,551 (+24h: +2,580, +48h: +3,690, +7d: +7,381, +30d: +17,812, +90d: +41,574). Observed after the launch; not proof that the launch caused the change.
1 related public event(s): reddit post. State before: About 59,517 stars the day before; repository 150 days old; 7 earlier launch episode(s).
Stars went from 59,517 the day before to 73,972 (+24h: +555, +48h: +774, +7d: +3,277, +30d: +14,455). Observed after the launch; not proof that the launch caused the change.
2 related public event(s): reddit post. State before: About 78,444 stars the day before; repository 197 days old; 8 earlier launch episode(s).
Stars went from 78,444 the day before to 95,095 (+24h: +461, +48h: +643, +7d: +3,691, +30d: +16,651). Observed after the launch; not proof that the launch caused the change.
1 related public event(s): reddit post. State before: About 81,685 stars the day before; repository 204 days old; 9 earlier launch episode(s).
Stars went from 81,685 the day before to 84,832 (+24h: +813, +48h: +1,129, +7d: +3,147). Observed after the launch; not proof that the launch caused the change.
1 related public event(s): reddit post. State before: About 88,824 stars the day before; repository 221 days old; 10 earlier launch episode(s).
Stars went from 88,824 the day before to 95,556 (+24h: +2,207, +48h: +3,178, +7d: +6,732). Observed after the launch; not proof that the launch caused the change.
1 related public event(s): reddit post. State before: About 93,515 stars the day before; repository 226 days old; 11 earlier launch episode(s).
Stars went from 93,515 the day before to 95,556 (+24h: +1,580, +48h: +2,041). Observed after the launch; not proof that the launch caused the change.
The repository was created on February 24, 2026. Releases followed quickly: v1.1.0 (Instagram, LinkedIn, Boss直聘) the next day, v1.2.0 (Douyin added, Instagram removed) on February 26, and v1.3.0 (WeChat 公众号 and Windows fixes) on March 4. The stated purpose is to give AI agents access to the internet with no API fees. The original README is in Chinese, with English, Japanese and Korean versions. The README says the author uses the tool daily. The claims do not say why the author started it beyond that.
The evidence does not identify specific early users. Some of the earliest audience may have been Chinese-speaking developers. The README was written in Chinese first, early releases added Chinese platforms such as Douyin and WeChat 公众号, and an AtomGit mirror exists for easier access from China. The repo had 5,618 stars about ten days after creation, according to a third-party tracker. Reddit posts quoting a tagline of 'One CLI, zero API fees' appeared from late March to early April 2026. It is not known how these visitors found the repo.
Weak inference| Aug 12 – Oct 11, 2026 | 2.6K forks |
| As of October 11, 2026 | 36 contributors |
| Aug 12 – Oct 11, 2026 | 7 pull requests |
| As of June 2026 | 42 issues | |
| Last push October 8, 2026 | 206 issues |
| Last 26 complete weeks | 134 commits |
| Date unknown | $1 |
| June 29, 2026 | 14 platforms | |
| As of October 5, 2026 | 16 platforms | |
| Date unknown | 16 platforms |
| Date unknown | 6 channels |
| Date unknown | 32 days |
Repeated appearances on trending lists, and a 'Repository of the Day' badge shown in the README.
How it works: Trending placement exposes the repo to new visitors, who add stars that sustain the ranking.
Third parties repeatedly reposted the tagline and star counts on Reddit and in articles.
How it works: A short, quotable pitch spreads across communities without action from the maintainer.
Instead of manual setup steps, users paste a single prompt containing the install.md URL into their AI agent, and the agent installs the tool itself.
undated: README instructs users to paste a one-sentence prompt pointing to the raw docs/install.md URL into Claude Code, Cursor or similar agents.
Package the tool so it can be added as a skill through an agent-skill installer, alongside pip.
undated: Installable via 'npx skills add' as well as 'pip install agent-reach'.
Lead with 'free, no API keys' against paid platform APIs, and disclose the only optional cost.
undated: Tagline 'one CLI, zero API fees'; README states all tools are free with only an optional ~$1/month proxy; uses cookie auth to access X instead of paying for the X API.
Model each platform as a channel with ordered backend choices, and ship a health-check command that reports which routes work and how to fix broken ones.
June 2026: Switched Bilibili from yt-dlp to bili-cli after yt-dlp was blocked; v1.5.0 added multi-backend routing and real health checks; 'agent-reach doctor' reports channel status.
Publicly remove channels that no longer work instead of advertising them.
June 10, 2026: Released v1.4.2, titled '诚实瘦身', which removed unmaintained channels and added transcription and doctor --json.
Publish the README in several languages and offer a mirror for a region where access to GitHub is harder.
undated: README is in Chinese, English, Japanese and Korean; an AtomGit mirror exists for easier access from China.
Ask users in the README to star the project, and state that the author uses the tool daily and will keep maintaining it.
undated: The README asks users to star the project.
Early releases added and swapped platforms: Instagram, LinkedIn and Boss直聘 in v1.1.0; Douyin added and Instagram removed in v1.2.0; WeChat 公众号 in v1.3.0. Reddit posts quoting the 'one CLI, zero API fees' tagline appeared in late March and early April.
In March 2026, after a batch of single-platform CLIs stopped being maintained, the project changed its routing. In June 2026 it removed channels that had stopped working, switched Bilibili from yt-dlp to bili-cli, and shipped v1.5.0, which added multi-backend routing, real health checks and OpenCLI. It frames itself as handling selection, installation, health checks and routing rather than being another tool.
From June through October, the repo repeatedly appeared on GitHub Trending and Trendshift. It was also covered by press (ToKnow.ai, Techbytes), posted on Hacker News, and listed in Reddit roundups.
The project describes itself as completely free, with only an optional ~$1/month server proxy. It lists sponsors (BrowserAct, Tencent Cloud OpenClaw, CoreClaw, UCloud AstraFlow), and the maintainer offers custom Agent implementation consulting via WeChat. A secondary source reports no paid tier and version 1.5.0 in August 2026.
Repository created Feb 24, 2026, followed by releases v1.1.0, v1.2.0 and v1.3.0 within about ten days; one secondary source reports 5,618 stars on March 5, 2026, roughly 10 days after creation.
Two choices seem to set the project apart. First, it does not write its own scrapers. It describes itself as a capability layer that selects, installs, health-checks and routes between existing free tools such as yt-dlp, gh, Jina Reader and Exa. Each platform has an ordered list of backends. When Bilibili blocked yt-dlp in June 2026, the Bilibili route was switched to bili-cli. A doctor command reports which channels currently work. Second, it removed broken channels openly: v1.4.2 was titled an 'honest slimming' that dropped unmaintained channels. Separately, onboarding is handed to the user's own agent through a one-sentence install prompt, and the tool is also distributed as an agent skill.
Moderate inference