OpenClaw AI Stats November 2026 – Uses, Users, Market Share, and More.

OpenClaw – the open-source AI agent formerly known as Clawdbot and Moltbot – went from one Austrian developer’s side project to the most-starred software project on GitHub in under a year.

It connects the big LLMs (Claude, GPT, Gemini, DeepSeek and dozens more) to the messaging apps people already use, like WhatsApp, Telegram and Slack, and then lets the agent get on with tasks.

The viral hype of early 2026 has settled into something more durable. GitHub star growth has largely flattened, but OpenClaw is now moving into the enterprise: Microsoft’s new Autopilot agent is built on it, and the OpenClaw Foundation has open-sourced an enterprise edition backed by OpenAI, Red Hat and Nvidia. Here are the most recent OpenClaw stats available as of November 2026.

OpenClaw acquired by OpenAI.

That was fast..

Peter bootstrapped OpenClaw, unprofitably, at a $10-20k/mo loss. This was his second acquisition. First exit for $116m, second undisclosed.

Build something people want.

Crazy things may happen. https://t.co/tV0jxaodRG

— JC (@shiftj) February 15, 2026

OpenClaw Stats – The Headlines (November 2026)

• OpenClaw has around 392,000 GitHub stars and 82,000 forks (October 2026) – still the most-starred software project on GitHub, though growth has largely flattened since the spring
• The openclaw npm package was downloaded around 13.5 million times in a single month (August 26 to September 24, 2026)
• OpenClaw.ai pulls in around 38 million monthly visitors, with an estimated 3.2 million monthly active users
• Microsoft’s Autopilot Copilot agent runs on OpenClaw, and OpenClaw Enterprise is now free and open-source
• The latest release, v2026.10.9 (October 29), focuses on security hardening
• The ClawHub skills marketplace has grown to 52,700+ extensions – but only 47% are rated safe
• 155,000+ OpenClaw instances are exposed on the internet without basic access protection

OpenClaw GitHub & Developer Stats

GitHub is where OpenClaw’s story started, and it’s still the best real-time measure of how much developers care about it.

~392,000 GitHub stars (October 2026). That’s up from 390,570 on September 26 (GitHub API figures reported by Gradually AI), alongside around 82,000 forks (The GenAI Magazine).

Takeaway: OpenClaw’s star count has barely moved since the spring – the viral phase is over, but nothing has come close to overtaking it.

It overtook React on March 3, 2026, at 250,829 stars. React needed more than a decade to reach that level; OpenClaw did it in about 60 days.

Takeaway: nothing in open-source history has been adopted by developers this quickly.

60,000 stars in 72 hours, and 190,000 in two weeks. That launch spike came in January 2026, after a Hacker News feature sent the project viral.

Takeaway: the early-2026 hype was real, but the bigger story is how much of that interest stuck around.

13.5 million npm downloads in one month. Between August 26 and September 24, 2026, the openclaw package was pulled from npm roughly 13.5 million times.

Takeaway: stars measure interest; downloads measure installs – and the install numbers are huge.

2,500 contributors in around seven months. That’s more than React gathered in ten years, with only giants like the Linux kernel and Kubernetes clearly ahead.

Takeaway: OpenClaw no longer depends on its founder – the community is now big enough to carry it.

OpenClaw 2.0 (v2026.8.1): 16,977 pull requests from 987 contributors in one release cycle. The release, announced August 30, 2026, added SQLite as the session store and a rebuilt browser app. The latest release at the time of writing is v2026.10.9 (October 29, 2026), a maintenance update focused on security hardening, reliability and performance (The Claw Report).

Takeaway: the project is shipping at a pace most commercial software teams couldn’t match.

OpenClaw Users & Web Traffic Stats

38 million monthly visitors. That’s the most recent Similarweb-based estimate for OpenClaw.ai, up from 27 million in March and just 7.25 million at the start of February 2026.

Takeaway: traffic has grown more than fivefold since February – though month-on-month growth is nowhere near the 925% spike seen between February and March.

3.2 million monthly active users. This is the latest available MAU estimate, up from the ~2 million figure that was circulating earlier in 2026.

Takeaway: a small user base next to ChatGPT or Gemini, but a large one for a self-hosted tool that requires technical setup.

45-minute average session duration and 92% monthly retention. Both figures come from Similarweb data and Gradually AI’s own research.

Takeaway: people who set OpenClaw up actually use it – sessions run almost twice as long as on ChatGPT.

20–30% of visits come from mobile. Most traffic to OpenClaw.ai is desktop, which makes sense given setup and admin happen on a computer.

Takeaway: the website is for setting it up; the day-to-day use happens on WhatsApp and Telegram.

Where OpenClaw Is Used

Country	Traffic Share	Monthly Change (at peak growth)
United States	16.29%	+611.21%
India	12.16%	+603.55%
China	12.08%	+1,436.45%
Germany	4.10%	+991.88%
Canada	3.53%	+1,259.18%

Takeaway: the US leads, but only just – the near three-way tie with India and China is unusual for a Western developer tool.

OpenClaw Model & Token Usage Stats

19.2 trillion tokens processed via OpenRouter since January 2026. That’s OpenClaw installations routing requests through OpenRouter alone, per OpenRouter’s app stats.

Takeaway: OpenClaw is one of the single largest consumers of LLM tokens on the planet.

Four of the six largest model providers used with OpenClaw are Chinese (StepFun, Xiaomi, Zhipu and MiniMax), together accounting for more than 54% of token usage. StepFun’s Step 3.5 Flash alone consumed 3.474 trillion tokens versus 2.135 trillion for Anthropic’s entire model line-up.

Takeaway: OpenClaw users optimise for cost – cheap, capable Chinese models are doing most of the heavy lifting.

356 different models have been used with OpenClaw. It supports OpenAI, Anthropic, Google, DeepSeek, Mistral and fully local models through Ollama.

Takeaway: being model-agnostic is OpenClaw’s biggest selling point – users aren’t locked in to any one AI lab.

OpenClaw Ecosystem & Business Stats

52,700+ skills on ClawHub, from 12,400 developers. The official skills marketplace went from zero to that total in under eight months, with 2.3 million skill installs counted.

Takeaway: ClawHub hit 10,000 extensions in four months – the VS Code Marketplace took over three years.

Only 6.8% of ClawHub publishers (847) are verified.

Takeaway: more than 93% of skills come from people who haven’t proven who they are – which feeds directly into the security problem below.

Tencent, Alibaba, ByteDance, Baidu and Xiaomi all use OpenClaw. In March 2026, more than 1,000 people queued outside Tencent’s Shenzhen HQ for installation help, and Mac Minis sold out across China.

Takeaway: China has embraced OpenClaw faster than any Western market, mostly paired with local models.

Peter Steinberger joined OpenAI in February 2026, and OpenClaw moved to a foundation. Sam Altman announced that OpenAI would keep supporting OpenClaw as an open-source project; the OpenClaw Foundation lists OpenAI among its major donors, alongside Vercel, Cloudflare and GitHub providing infrastructure.

Takeaway: OpenClaw remains MIT-licensed and independent on paper, but OpenAI now sits closer to it than any other lab.

Microsoft’s Autopilot runs on OpenClaw, and OpenClaw Enterprise is free. Microsoft shipped Autopilot, an always-on Copilot agent built on OpenClaw, and the OpenClaw Foundation open-sourced OpenClaw Enterprise – an MIT-licensed control plane that adds multi-tenancy, audit trails and permission controls, with OpenAI and Red Hat piloting it internally (OpenClaw community blog; BigGo Finance).

Takeaway: OpenClaw is moving from developers’ side projects into the enterprise stack.

OpenClaw Security Stats

36% of ClawHub skills contain prompt injections. Analysis cited by Gradually AI from Cisco Talos and Kaspersky found just 47% of skills were safe, 8% tried to send user data to external servers and 6% requested excessive permissions.

Takeaway: installing a random skill is a genuine risk – review the source code first.

155,000+ unprotected OpenClaw instances are reachable online. That’s roughly one in fifteen installations, according to the same research.

Takeaway: most of OpenClaw’s security risk comes from how people deploy it, not just the code.

Two new CVEs in September 2026. CVE-2026-95815 (bearer tokens logged by the iOS app, fixed in v2026.8.11) and CVE-2026-94094 (a denial-of-service flaw, still unresolved per the advisory).

Takeaway: vulnerability reports haven’t slowed down – keeping OpenClaw updated isn’t optional.

Summing Up OpenClaw Statistics (November 2026)

OpenClaw is still a niche product next to ChatGPT, Google Gemini or even Perplexity in raw user numbers, and its GitHub growth has flattened.

But as of November 2026, it’s still the most-starred project on GitHub, one of the biggest token consumers in AI, and – with Microsoft’s Autopilot and OpenClaw Enterprise – increasingly the open-source foundation that big tech builds agents on. The question for 2027 is whether the security work in recent releases can keep pace with that enterprise ambition, as newer agents like Meta Muse bring agents to the mainstream.

For the bigger picture, see our AI Usage Statistics round-up.

Sources: The GenAI Magazine (Oct 2026) · The Claw Report · OpenClaw community blog · BigGo Finance · Gradually AI – OpenClaw Statistics (Sep 30, 2026) · GitHub – openclaw/openclaw · OpenRouter · Sam Altman on X · NVD