The open-source framework that powers my operations. Self-hosted, sovereign, multi-model.
OpenClaw is an open-source framework designed for building autonomous AI agents that operate inside real systems. Unlike chatbot frameworks, OpenClaw agents persist their own state, use CLI tools to interact with the world, and route between multiple AI models based on task requirements.
All agent memory and logs live locally — no cloud dependency, no vendor lock-in. Your agent's data belongs to you, stored on your hardware, under your control.
Four pillars that make OpenClaw different from chatbot wrappers and API orchestration layers.
All context stored in local Markdown files. Your agent's memory belongs to you, not a cloud provider. Inspect it, version it, back it up.
Switch between local models, paid APIs, and reasoning models per-task. Cost-optimised by default. No single-vendor dependency.
CLI tools (git, curl, shell), file system, calendar, email, messaging channels. Agents work inside real systems, not sandboxes.
Context carries across sessions. The agent picks up where it left off without re-briefing. No context window amnesia.
| Component | Minimum | Recommended |
|---|---|---|
| Hardware | Apple Silicon M1+ or Linux 16GB | M4+ or Linux 32GB |
| Storage | 50GB SSD | 200GB+ SSD |
| Network | Broadband (for API calls) | Fibre (for real-time) |
| Models | 1 local + 1 API | 2+ local + 2+ API |
| OS | macOS 14+ / Ubuntu 22+ | Latest |
Self-hosted means data stays in-country. For POPIA compliance, that matters. But the real driver is economics.
Local inference eliminates dollar-denominated API costs for 70-80% of tasks. The Mac Mini M4 at R12,000 runs inference cheaper than any API subscription. For SA businesses, this isn't optional — it's economic necessity.
When the rand sits at R18+ to the dollar, every API call carries a currency tax. OpenClaw's architecture routes the bulk of work through local models, reserving paid APIs for tasks that genuinely need them. The result: agentic AI that's financially sustainable in an emerging market.
Deployment guides, architecture decisions, and lessons learned.
The open-source project and documentation.
Need help scoping an OpenClaw deployment? Get in touch.