The short version
OpenClaw is a free, open-source AI assistant that you run on your own computer or server instead of renting access from a big tech platform. Once it is running, you talk to it through the chat apps you already use, such as WhatsApp, Telegram, Slack, or Signal. You ask for things in plain language, and it goes off and does them.
That last part is the key difference from an ordinary chatbot. OpenClaw is an agent. It does not just answer questions; it can read files, browse the web, run scripts, watch inboxes, and carry out multi-step tasks on your behalf. The official repository at github.com/openclaw/openclaw sums it up with the tagline "Your own personal AI assistant. Any OS. Any Platform. The lobster way."
The project is MIT-licensed, written mostly in TypeScript and Swift, and has become one of the fastest-growing open-source projects in GitHub history, passing 300,000 stars by mid 2026. In Mauritius we have seen the same curiosity as everywhere else: business owners hearing about it in WhatsApp groups and wondering whether it can take real work off their plate.
Bring your own model
OpenClaw does not ship with its own AI brain. It is model-agnostic, meaning you connect it to a large language model of your choice, such as Anthropic's Claude, OpenAI's GPT models, or DeepSeek, using your own API keys.
This matters for two reasons. First, you control the costs: you pay the model provider directly for what you use, with no markup. Second, you control the data path: your conversations flow between your hardware and your chosen provider, with no extra middleman. The official documentation at docs.openclaw.ai recommends using a current flagship model, because older and smaller models handle tools and hostile input noticeably worse.
How it works, at a high level
Three ideas cover most of the architecture:
- The gateway: a small control plane that runs on your machine and connects your chat channels to the model. It is the front door, and how you secure it matters a great deal.
- Skills: folders containing a SKILL.md file that teaches the agent how to do a specific job, such as auditing a website or summarizing invoices. Skills can be bundled, installed globally, or scoped to a single workspace.
- Plugins: extensions to the runtime itself, used for deeper integrations.
Skills are shared through a community marketplace called ClawHub. It is convenient, but as we cover elsewhere on this site, it has also been a target for attackers, so nothing from it should be installed blindly.
A short, slightly chaotic history
OpenClaw has had five names in just over two months. Austrian developer Peter Steinberger released it as Warelay in November 2025. It became CLAWDIS in December, then Clawdbot in early January 2026, then Moltbot at the end of that month after trademark complaints from Anthropic, and finally OpenClaw on 30 January 2026.
In February 2026 Steinberger announced he was joining OpenAI to work on personal agents, with an independent non-profit, the OpenClaw Foundation, set up to steward the open-source project. Releases now follow a calendar scheme such as v2026.6.11, and the June 2026 line focused on reliability fixes and safer admin defaults.
Why we recommend caution before self-running
Everything above is genuinely impressive, and this site exists because we think the technology is worth understanding. But an agent that can read your email, use your credentials, and run commands on your machine is a serious piece of infrastructure, not a toy.
OpenClaw's own security documentation is candid: it is designed as a single-user personal assistant and trusts the operator to lock it down. Through 2026 there has been a steady stream of vulnerability disclosures, research on exposed instances, and a supply-chain attack on the skill marketplace. None of that makes OpenClaw a scam. It makes it powerful software with sharp edges.
Our advice to Mauritian readers is simple: read the official docs, explore the repo if you are curious, and understand the risks before you connect anything you care about. If you want the benefits for your business without becoming a part-time security engineer, a hardened setup by the local specialists at nexus.mu is the shortcut we actually recommend. Either way, start informed. The rest of this site exists for exactly that.
Powerful agents deserve professional setup, not blind cloning. Explore the wider Nexus health ecosystem.
