AI desktop agent that sees your screen, controls your cursor, and completes tasks autonomously. Features a 5-layer intelligent fallback pipeline, multiple AI providers (Anthropic/OpenAI/Ollama/Kimi), with Web Dashboard and REST API.
Clawd Cursor is an open-source AI desktop agent that enables AI to "see your screen, control your cursor, and complete tasks." It achieves cross-application automation through a 5-layer intelligent pipeline.
Key Features#
- 5-Layer Fallback Pipeline: Browser (CDP) → Action Router → Smart Interaction → Accessibility Reasoner → Screenshot + Vision; cost-effective first, graceful degradation
- Multiple AI Providers: Anthropic (recommended for complex tasks), OpenAI, Ollama (local/free), Kimi
- Self-Healing Doctor: Auto-detects environment, tests models, builds optimal pipeline
- Web Dashboard: Browser UI (localhost:3847) for task submission, real-time logs, safety confirmations, kill switch
- Safety Tiers: Auto (immediate) → Preview (log then execute) → Confirm (user approval required)
- Privacy Protection: REST API binds to localhost only; Ollama mode runs fully offline
Use Cases#
- Cross-app data migration (web → desktop app)
- Form filling and web navigation
- UI verification and state checking
- Batch action execution
- Automated testing/demos
Quick Start#
git clone https://github.com/AmrDab/clawd-cursor.git
cd clawd-cursor
npm install && npm run build
npm run doctor
npm start
Platform Support#
- Windows/macOS: Full functionality
- Linux: Browser layer only
Prerequisites#
- Node.js >= 20
- macOS requires Terminal Accessibility permission