An open-source AI assistant called OpenClaw has hit 350,000 GitHub stars and is being downloaded half a million times a day. If you haven't heard of it yet, you will.
For Australian businesses — particularly those in regulated industries or with any sensitivity around where their data lives — it's worth understanding why this one is different from the AI tools that came before it.
What OpenClaw Is
OpenClaw is a self-hosted, open-source AI agent that runs on your own infrastructure and communicates through the messaging platforms your team already uses: Slack, WhatsApp, Telegram, Discord, Teams, Google Chat, Signal, iMessage, and others.
You own the installation. You own the data. The AI runs on your hardware, calls the LLM of your choice — local models, OpenAI, Anthropic, or anything OpenAI-compatible — and responds through whatever channel you configure.
That architecture is the whole point. Most AI assistants are cloud products where your conversations, prompts, and business data are sent to someone else's servers. OpenClaw flips that model entirely.
A Brief History
OpenClaw was first published in November 2025 by Austrian developer Peter Steinberger under the name ClawdBot. After a trademark dispute with Anthropic, it was briefly renamed Moltbot before settling on OpenClaw in January 2026. The rebrand did nothing to slow adoption — if anything the controversy accelerated it.
By April 2026, it had become one of the fastest-growing open-source projects in history.
How It Works
Once deployed, OpenClaw acts as a persistent AI agent connected to your business systems. A typical setup looks like this:
- You self-host the OpenClaw server on your own VPS, on-premise machine, or private cloud
- You connect it to your messaging platform of choice (most teams use Slack)
- You configure which LLM it calls — a local model for sensitive data, or a cloud API for general tasks
- You install skills — modular plugins that extend what the agent can do
The marketplace currently has over 13,700 skills covering everything from Google Ads management and CRM integration to calendar scheduling, document summarisation, and competitive monitoring.
Why It Resonates in Australia
Australian businesses have been slower than their US counterparts to fully commit to cloud AI tools — and the reason is usually data. Whether it's the Privacy Act, industry-specific compliance requirements, or simply a well-founded reluctance to send client data offshore, the question of where does this data actually go comes up constantly.
OpenClaw's local-first architecture is a direct answer to that question. When the model runs on your server and never phones home, data residency stops being a problem.
This matters most for:
- Healthcare — patient data stays within your practice's infrastructure
- Financial services — client conversations and financial records never leave your environment
- Legal — privileged communications remain genuinely privileged
- Government and education — compliance with data handling obligations without architectural compromises
What It Costs
The software itself is free and open-source. Real costs are infrastructure:
| Component | Approximate Monthly Cost |
|---|---|
| VPS for self-hosting | $20–200 AUD depending on size |
| Local LLM inference (GPU instance) | $100–500 AUD for heavier workloads |
| Cloud LLM API calls (if used) | Variable — pay per token |
| Most OpenClaw skills | Free |
| Premium skills | Varies, most under $50 AUD one-time |
For a small business running modest automation, total infrastructure cost can be under $100/month AUD — a fraction of what a comparable cloud AI assistant subscription would run.
Practical Use Cases
Daily briefings — OpenClaw pings your team on Slack each morning with a summary pulled from your CRM, calendar, and project tracker. No manual collation.
Customer support triage — incoming support requests get classified, prioritised, and routed to the right person. Simple queries get answered automatically.
Internal knowledge assistant — point OpenClaw at your documentation, SOPs, and knowledge base. Staff ask questions in Slack and get accurate answers drawn from your own material.
Sales and CRM automation — new leads trigger enrichment, follow-up drafts, and CRM updates without manual data entry.
Reporting — weekly performance reports assembled and delivered automatically to whoever needs them.
OpenClaw vs Cloud AI Assistants
The comparison most teams make is against tools like Microsoft Copilot, Google Gemini for Workspace, or ChatGPT Enterprise.
| OpenClaw | Cloud AI Assistants | |
|---|---|---|
| Data location | Your infrastructure | Provider's cloud |
| Monthly cost | Infrastructure only | Per-seat subscription |
| Customisation | Full (open source) | Limited to platform settings |
| Skills/integrations | 13,700+ community skills | Curated by vendor |
| Setup complexity | Moderate | Low |
| Data sovereignty | Complete | Dependent on provider |
The honest tradeoff: cloud AI assistants are easier to set up. OpenClaw requires a deployment. For most businesses with any sensitivity around data or cost-at-scale, the tradeoff is worth it.
Getting Started
The practical starting point for most businesses is a scoped pilot:
- Identify one or two repetitive, high-frequency tasks where an AI agent would save time
- Deploy OpenClaw on a small VPS — the setup process has improved significantly with recent releases
- Connect to Slack or whichever messaging platform your team lives in
- Install the relevant skills and test against your actual workflows
The most common entry points we see are internal knowledge assistants and daily reporting automation — both are low-risk, immediately useful, and give teams a realistic sense of what OpenClaw can do before expanding scope.
If you want to talk through what an OpenClaw deployment looks like for your specific business, get in touch. We work with Brisbane and Australian businesses on AI infrastructure from architecture through to production deployment.
Sources: GitHub — openclaw/openclaw · OpenClaw Wikipedia · KDnuggets · Linux Journal