EClaw is a safe embedded Claw Bot for Raspberry Pi 5 and basic Mac Mini, derived from Openclaw. It retains core capabilities—multi-message channels, vector memory, browser automation, canvas interaction, voice, and hybrid memory—while simplifying the codebase and architecture to improve maintainability and development speed.
Delivering most of the capability of large agent platforms at a fraction of the complexity.
EClaw sits between large, enterprise agent systems and ultra-minimal personal assistants:
Capability
↑
│ OpenClaw
│ (Enterprise platform)
│ EClaw
│ (Balanced, practical)
│
│
│
│
│ NanoClaw
│ (Minimal personal)
└──────── Low Complexity ───────── High Complexity →
This balance allows EClaw to provide real-world functionality while remaining maintainable and easy to understand.
| Dimension | OpenClaw | EClaw | NanoClaw |
|---|---|---|---|
| Philosophy | Enterprise-grade platform | Balanced middle ground | Minimalist personal tool |
| Lines of Code | ~585K TS + Swift/Kotlin | ~33K TS | ~4.3K TS |
| Source Files | 2,232 TS + 438 Swift + 66 Kotlin | 217 TS | 22 TS |
| Messaging Channels | 7 core + 30 extensions | 7+ channels + Web UI | 1 (WhatsApp only) |
| LLM Providers | 20 providers | Anthropic + OpenRouter | Claude via Agent SDK |
| Mobile Apps | macOS, iOS, Android | None | None |
| Test Coverage | 70%+ enforced | Vitest + Playwright | None |
| Dependencies | 52 runtime | 20 runtime | 7 runtime |
git clone https://github.com/wubo3x/eclaw.git
cd eclaw
claudeEClaw is designed around three core principles:
- Multi-channel messaging
- Hybrid memory
- Browser automation
- Canvas interaction
- Voice and tools
- Scheduled tasks
- Smaller codebase
- Fewer dependencies
- Clear modular structure
- Easy debugging
- Fast setup
- Simple configuration
- Hot reload
- Graceful degradation
- Independent developers building agent tools
- Small teams deploying assistants
- Developers who prefer readable code
- Projects needing multi-channel automation
EClaw aims to strike a balance between capability and simplicity:
- Powerful enough for real-world use
- Small enough to understand
- Flexible enough to extend
- Stable enough to maintain
MIT