Desktop/browser automation via ClawdCursor API. Send natural language tasks to control browser and desktop apps.
AI desktop agent — control browser and apps via natural language tasks.
# Start ClawdCursor (uses Ollama + Playwright)
bash ~/clawd/skills/clawd-cursor/start.sh
| Feature | Linux (headless) | Windows | macOS | |---------|-----------------|---------|-------| | Browser automation (Playwright) | ✅ Full | ✅ Full | ✅ Full | | URL navigation | ✅ | ✅ | ✅ | | Form filling | ✅ | ✅ | ✅ | | Screenshots | ✅ | ✅ | ✅ | | Native desktop (mouse/keyboard) | ⚠️ Xvfb | ✅ | ✅ | | Accessibility tree reasoning | ❌ | ✅ | ✅ | | LLM planning (Ollama) | ✅ | ✅ | ✅ |
Linux Note: Browser automation via Playwright works perfectly. Native desktop (Start menu, app switching) works in Xvfb but has limited accessibility APIs.
ClawdCursor runs at http://127.0.0.1:3847:
curl -s -X POST http://127.0.0.1:3847/task \
-H "Content-Type: application/json" \
-d '{"task": "Navigate to https://github.com"}'
curl -s http://127.0.0.1:3847/status
curl -s -X POST http://127.0.0.1:3847/confirm \
-H "Content-Type: application/json" \
-d '{"approved": true}'
curl -s -X POST http://127.0.0.1:3847/abort
| Task | Example |
|------|---------|
| Navigate | Navigate to https://example.com |
| Search | Go to google.com and search for AI news |
| Fill form | Go to example.com/form and fill name=John, email=john@test.com |
| Click element | Click the Sign In button |
| Type text | Type "hello world" in the search box |
| Task | Example |
|------|---------|
| Launch app | Open Chrome |
| Keyboard | Press Ctrl+C |
| Window management | Minimize all windows |
Layer 0: Browser (Playwright) → URL detection → instant ✅
Layer 1: Action Router (regex) → Pattern matching → instant ✅
Layer 1.5: Smart Interaction → CDP + UI automation + 1 LLM call
Layer 2: A11y Reasoner (Ollama) → Text-only LLM reasoning
Layer 3: Vision (requires API key) → Screenshot → vision LLM
On Linux headless: Layers 0-1 work best. Layer 1.5+ have limited accessibility support.
ClawdCursor uses Ollama for LLM reasoning:
# Verify Ollama is running
curl -s http://localhost:11434/api/version
# Model: qwen2.5:3b (pre-configured)
OLLAMA_MODELS=/data/ollama/models /data/ollama/bin/ollama list
Config stored in: ~/clawd/clawd-cursor/.clawd-config.json
ClawdCursor complements ClawdBot's built-in browser tool:
Use ClawdCursor when you want to describe a goal in natural language rather than individual actions.
curl -s http://127.0.0.1:3847/status
tail -f /tmp/clawd-cursor.log
pkill -f "clawd-cursor"
bash ~/clawd/skills/clawd-cursor/start.sh
pkill -f "node dist/index.js"Problem: ClawdCursor needs a display, but headless Linux has none.
Solution: Use xvfb-run --auto-servernum --server-args="-screen 0 1920x1080x24" to create a virtual framebuffer.
Gotcha: Accessibility APIs are limited in Xvfb — Layer 2+ reasoning may not work as well as on real displays with full a11y support. Browser automation (Playwright) works perfectly though.
Pattern: ClawdCursor uses local Ollama for LLM reasoning. On resource-constrained systems:
OLLAMA_MODELS to a data partition with spaceLearning: ClawdCursor works best with goal-oriented tasks, not micro-instructions.
The LLM planner breaks down goals into steps — let it do its job.
npx skills add OperatingSystem-1/clawd-cursor下载完整 Skill 目录，包含 SKILL.md 及所有相关文件
Search for places (restaurants, cafes, etc.) via Google Places API proxy on localhost.
Interact with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
Create or update AgentSkills. Use when designing, structuring, or packaging skills with scripts, references, and assets.
Start voice calls via the OpenClaw voice-call plugin.
Notion API for creating and managing pages, databases, and blocks.
Gemini CLI for one-shot Q&A, summaries, and generation.
Category:developer