Getting Started
Get your operator cockpit connected to OpenClaw in 5 minutes.
Claw Desktop is the operator cockpit for OpenClaw agents. Your agent runs remotely—on DigitalOcean, Cloudflare, or your own server—working overnight while you chat via Slack. When you’re back at your desk, open Claw Desktop to review what happened, check the diffs, and resume exactly where your agent left off.
Why Claw Desktop?
Stop scrolling through Slack history to find what your agent did overnight. Claw Desktop gives you:
- While-You-Were-Away Sync — See exactly what changed since you last opened the app
- Artifact Review — Diffs, outputs, and verification checklists in one place
- One-Click Resume — Continue the same
run_idacross Slack and Desktop - Human-in-the-Loop Approvals — Approve or reject blocked actions inline
Download
| Platform | Download |
|---|---|
| macOS (Apple Silicon) | Download .dmg |
| macOS (Intel) | Download .dmg |
| Windows | Download .exe |
Or grab it from the releases page.
Connect to Your OpenClaw Instance
Claw Desktop connects to your OpenClaw instance via WebSocket. You’ll need:
- Host URL — Where your OpenClaw is running (e.g.,
ws://your-server:8080) - API Token — Your OpenClaw API token for authentication
Setup Steps
- Launch Claw Desktop
- Enter your Host URL and API Token
- Click Test Connection to verify
- Save your connection settings
You can add multiple OpenClaw hosts and switch between them.
How It Works
┌─────────────────┐ ┌──────────────────┐ ┌─────────────────┐│ Your Slack │◄────►│ OpenClaw │◄────►│ Claw Desktop ││ (Chat while │ │ (Runs 24/7 on │ │ (Review when ││ you sleep) │ │ your server) │ │ you're back) │└─────────────────┘ └──────────────────┘ └─────────────────┘- Deploy OpenClaw on DigitalOcean, Cloudflare, or your infrastructure
- Chat via Slack to kick off runs and monitor progress
- Open Claw Desktop to review artifacts, approve blocked actions, and resume
Your run_id stays consistent across all interfaces. Context is never lost.
Core Features
Runs Inbox
See all your runs at a glance:
- Status chips — Running (blue), Blocked (yellow), Done (green), Failed (red)
- Filter tabs — All, Running, Blocked, Done
- While You Were Away — Updates since your last session, newest first
Run Timeline
Chronological view of everything that happened:
- Messages — User and assistant messages with timestamps
- Tool Calls — Collapsed cards showing tool name, expandable for params/results
- File Diffs — Syntax-highlighted patches showing exactly what changed
Artifact Panel
Right panel with three tabs:
- Plan — The agent’s plan (PLAN.md)
- Summary — What was accomplished (SUMMARY.md)
- Verify — Verification checklist (VERIFICATION.md)
Human-in-the-Loop
When your agent is blocked on an action:
- See the blocked action in the timeline
- Click Approve or Reject inline
- Run continues automatically after your decision
Architecture
Claw Desktop follows a 4-tier architecture:
┌───────────────────────────────────────────────────────────┐│ UI Layer ││ (SolidJS Components + Reactive Stores) │├───────────────────────────────────────────────────────────┤│ Sync Engine ││ (Cursor-based polling every 5 seconds) │├───────────────────────────────────────────────────────────┤│ Persistence ││ (File-system: ~/Documents/ClawWorkspace/RUNS/<id>/) │├───────────────────────────────────────────────────────────┤│ OpenClaw Adapter ││ (WebSocket V3 Protocol Translation) │└───────────────────────────────────────────────────────────┘File-System Persistence
All run data is stored locally in ~/Documents/ClawWorkspace/:
ClawWorkspace/└── RUNS/ └── <run_id>/ ├── meta.json # Run metadata ├── events.jsonl # Append-only event log └── ARTIFACTS/ # Downloaded artifactsYou can browse this workspace in Finder without the app.
Get Help
- Documentation — Browse the sidebar for detailed guides
- Discord — Join the community
- GitHub — ClawHQ/claw
What’s Next?
- Connecting to OpenClaw — Detailed connection setup
- Runs Inbox — Managing your runs
- Human-in-the-Loop — Approval workflows