OpenClaw: the open-source AI assistant that changed what people expect
A developer in Austria named Peter Steinberger built an AI assistant you run on your own hardware. It reaches you on WhatsApp, Telegram, iMessage, Discord, and about 20 other apps you already use. No subscription. No company holding your data. It hit 335k stars on GitHub faster than almost any open-source project in history. Then the big players started building toward the same idea.
What it is
OpenClaw is a personal AI assistant you install and run yourself. Once it’s set up, it lives across whichever messaging apps you already use and responds there. It supports voice on iOS, Android, and macOS, works with multiple AI models, and keeps memory of your conversations across sessions. You pay for the hosting and the model API, roughly €5-15 a month depending on how much you use it, and that’s it.
It started life in November 2025 under the name Clawdbot, got renamed to Moltbot after Anthropic raised trademark concerns, and then renamed again three days later to OpenClaw because, as Steinberger put it, Moltbot never quite rolled off the tongue.
Who’s using it and how
Right now it’s mostly developers and technically minded people who are comfortable setting up a Linux server or running a Node.js service. A lot of early adopters bought a Mac Mini specifically to keep it running around the clock, which tells you something about how much some people wanted this kind of setup.
That said, enterprise adoption has been picking up faster than most expected. By early 2026, one AI solutions provider reported that about a third of their new enterprise customers were actively moving away from managed cloud agent services and onto self-hosted OpenClaw infrastructure. The appeal is straightforward: your data stays on your own machines, and you’re not locked into any single AI provider.
What it proved
OpenClaw showed real demand for something mainstream AI products weren’t offering: an always-on assistant with long-term memory, multi-channel reach, and no lock-in.
When Anthropic shipped the ability to message Claude Code via Telegram and Discord with sessions that persist over time, people noticed the resemblance straight away. One comment doing the rounds summed it up: “They’ve built OpenClaw.” That’s not a coincidence, it’s a signal about where things are heading.
What’s happened since
In February 2026, Steinberger announced he was joining OpenAI. Fork activity on GitHub jumped 400% in 24 hours as people worried the project might be abandoned, but that didn’t happen. He handed leadership to a seven-person steering committee made up of experienced maintainers, and development has continued.
Security has been the main ongoing concern. Cisco’s research team found a third-party OpenClaw skill that was performing data exfiltration and prompt injection without users knowing. Around 1,100 security advisories have been filed since launch, though roughly 650 have been resolved. In March 2026, China restricted state agencies and state-owned enterprises from running it, citing security risks. It’s not a reason to avoid OpenClaw, but it is a reason to think carefully about how you set it up.
The ownership question
This is where it gets interesting, especially if you’re thinking about AI beyond the headlines.
OpenClaw sits at one end of a spectrum: full control, full responsibility, open source. The big proprietary tools sit at the other end: easier, more polished, but running on someone else’s infrastructure, under someone else’s terms. Most people will end up somewhere in the middle without having thought much about it. The choice of where your AI assistant lives, who can access your conversations, and what happens to that data over time isn’t one most of us are making consciously yet. Probably worth starting to.
And then, eight weeks later, there was Hermes
This is probably the best illustration of how fast this space moves. OpenClaw spent months becoming the go-to self-hosted AI assistant, built up an ecosystem of hundreds of thousands of users, and then in February 2026 a research lab called Nous Research dropped something that immediately started eating into that conversation.
Hermes is open-source, self-hosted, works across the same messaging apps, and costs roughly the same to run. On the surface it looks like an OpenClaw clone. The difference is one specific thing: it learns. Every task it completes, every problem it solves, it quietly writes a small file capturing what worked. Next time a similar task comes up, it pulls that in. After a few months of daily use, benchmarks show it completing similar tasks around 40% faster than a fresh instance would. OpenClaw doesn’t do this.
Nous Research also made the migration from OpenClaw deliberately easy, just one command that imports your settings, memories, and skills. When Anthropic blocked OpenClaw from accessing Claude Code subscriptions in early April, Nous Research tweeted: “If you’re having trouble with your lobster-themed agent since the recent update, try downloading Hermes Agent, then running ‘hermes claw migrate’.” It picked up 800 likes in a few hours and migration traffic spiked immediately.
Hermes isn’t ready for everyone yet. Documentation is thin, the API changes between releases, and the self-learning feature is turned off by default. But it crossed 100,000 GitHub stars in about seven weeks.
The point isn’t that Hermes beats OpenClaw. It’s that eight weeks after launch it’s already a serious conversation. Whatever is the dominant personal AI agent today probably won’t be in six months, and the reasons it gets replaced will likely have nothing to do with marketing budgets or brand recognition. Someone will just build something that does one thing noticeably better, make migration easy, and the community will move.
If you could have an AI assistant that runs entirely on your own hardware and reaches you on any app you already use, what's the first thing you'd actually use it for? Tell me in the comments 👇👇
Sources
OpenClaw GitHub repository: github.com/openclaw/openclaw
OpenClaw official website and documentation: openclaw.ai / docs.openclaw.ai
DigitalOcean: What is OpenClaw? digitalocean.com/resources/articles/what-is-openclaw
KDnuggets: OpenClaw Explained, March 2026: kdnuggets.com/openclaw-explained-the-free-ai-agent-tool-going-viral-already-in-2026
Technerdo: What Is OpenClaw? April 2026: technerdo.com/post/what-is-openclaw
Hermes Agent GitHub repository: github.com/NousResearch/hermes-agent
Hermes Agent official website: hermes-agent.nousresearch.com
VentureBeat: Anthropic ships Claude Code Channels: venturebeat.com/orchestration/anthropic-just-shipped-an-openclaw-killer-called-claude-code-channels

