OpenClaw AI Hits 150K GitHub Stars: The Self-Hosted Personal AI Guide
The short answer: OpenClaw is a free, open-source AI agent that runs on your own computer or server and takes instructions through chat apps like WhatsApp, Telegram, and Discord. The "150K" figure refers to GitHub stars, not registered members. The project passed it roughly ten weeks after launch, and it has grown far beyond that since.
This guide covers what OpenClaw does, why it grew so fast, how self-hosting compares with cloud assistants, and what to check before giving an AI agent access to your machine.
What Is OpenClaw?
OpenClaw is an open-source personal AI assistant created by Peter Steinberger, an Austrian engineer who founded PSPDFKit. It is released under the MIT license. Unlike a chatbot that only answers questions, it is built to carry out tasks: handling email, managing files, automating a browser, and running terminal commands.
You talk to it through messaging apps you already use. Supported channels include WhatsApp, Telegram, Discord, Slack, Teams, Signal, and iMessage. It is also model-agnostic, so you bring your own API keys for models such as Claude or GPT, or connect local models through Ollama.
The Name History
The project launched in late November 2025 as Clawdbot. After a trademark request from Anthropic on January 27, 2026 (the name was too close to "Claude"), it was renamed Moltbot and then settled on OpenClaw. If you see older articles under those names, they describe the same project.
What the 150K Milestone Actually Means
Important: some headlines describe the number as "members." The sources I could verify point to GitHub stars. One analysis reported more than 150,000 stars, over 22,000 forks, and a very active Discord community about ten weeks after launch. Stars are a popularity signal, not a user count, so they should not be read as 150,000 active users.
| Period | Reported GitHub stars | Notes |
|---|---|---|
| About 10 weeks after launch | 150,000+ | 22,000+ forks reported |
| Early March 2026 | About 245,000 | Per a third-party statistics tracker |
| April 8, 2026 | About 350,000 | About 1,600 contributors |
| Most recent figures | About 375,000 | About 78,000 forks |
Star counts change daily, so check the official OpenClaw repository for the current number.
Why OpenClaw Grew So Fast
- It acts, not just answers. People wanted an assistant that finishes tasks instead of suggesting steps.
- It lives in your chat apps. There is no new interface to learn.
- You own the setup. It runs on a laptop, homelab, or VPS you control, with your keys and your data location.
- It is free and MIT-licensed. You pay only for the model API usage and any hosting.
- The community built around it. A Discord community and projects like Moltbook, a social network for AI agents, kept attention high.
Self-Hosted vs Cloud AI Assistants
| Factor | Self-hosted (OpenClaw) | Cloud assistant |
|---|---|---|
| Data location | Your machine or server | Provider's servers |
| Cost | Free software, plus API and hosting costs | Usually a subscription |
| Setup effort | Higher, technical | Minimal |
| Model choice | Any supported model | Provider's models only |
| Security responsibility | Mostly yours | Mostly the provider's |
Self-hosting gives you control, but it also makes you the administrator. Note that your prompts still go to whichever model provider you connect, unless you use a local model.
Who Should (and Shouldn't) Use It
A good fit if you:
- Are comfortable with a terminal, API keys, and basic server administration.
- Want automation across email, files, and messaging in one place.
- Care about choosing your own models and hosting.
Probably not yet if you:
- Want a zero-maintenance, plug-and-play assistant.
- Would run it on your main computer with sensitive accounts logged in and no safeguards.
- Don't want to monitor API spending.
Security: The Part Most Guides Skip
The key point: an agent with access to your files, browser, and terminal, and that also reads untrusted content like emails and web pages, is risky by design. A malicious message could try to trick the agent into doing something you didn't intend. This is a known issue for agents generally, not just OpenClaw.
Sensible precautions:
- Isolate it. Use a dedicated machine, VM, or VPS rather than your daily computer.
- Limit access. Give it only the accounts and folders it needs.
- Don't expose it publicly. Restrict who can message the bot and keep management interfaces off the open internet.
- Vet add-ons. Review any third-party skills or plugins before installing them.
- Cap spending. Set usage limits on your model API keys.
- Keep it updated. Follow the project's release notes for security fixes.
How to Get Started
Follow the installation steps in the official repository and documentation rather than third-party tutorials, since commands and configuration change between releases. In general, the process looks like this:
- Choose where it will run: a spare computer, home server, or VPS.
- Install it using the official instructions.
- Connect one messaging channel first, such as Telegram, and confirm it works.
- Add your model API key, or point it at a local model.
- Start with low-risk tasks like summaries and reminders before granting email or file access.
Frequently Asked Questions
Did OpenClaw really reach 150K members?
It reached more than 150,000 GitHub stars within about ten weeks of launching. That measures interest, not active users, and the star count has since grown well beyond that.
Is OpenClaw free?
The software is free and MIT-licensed. You still pay for model API usage and for hosting if you rent a server.
Is OpenClaw the same as Clawdbot?
Yes. It was originally called Clawdbot, briefly Moltbot, and is now OpenClaw.
Can I use it without sending data to a cloud AI provider?
Yes, if you connect a local model through a tool like Ollama. With a cloud model, your prompts go to that provider.
Is it safe to run on my main computer?
It is safer to run it on an isolated machine or server with limited permissions, because the agent can act on whatever it can access.
Final Takeaway
OpenClaw's growth shows real demand for AI assistants that run on infrastructure you control and actually complete tasks. That power comes with responsibility: set it up in isolation, limit its permissions, and watch your costs. If you're ready for that, it's one of the most capable open-source options available. Next, you can read a related guide or compare it with other self-hosted AI tools.