If you're running an OpenClaw AI agent, you've probably hit the point where managing everything through chat feels like herding cats. Tasks pile up, you lose track of token usage, and editing your agent's personality means SSH-ing in and hand-editing markdown files.
VidClaw fixes this. It's an open-source, self-hosted dashboard that gives you a visual command center for your OpenClaw agent — kanban boards, usage tracking, model switching, and more. Here's how to get it running in under five minutes.
OpenClaw is the engine that powers your AI agent — it handles scheduling, memory, tool use, and model communication. But it runs headlessly. That means you're left juggling terminal sessions, config files, and chat interfaces to manage what your agent does.
A dashboard layer like VidClaw gives you:
vimAll of this runs on your own server, bound to localhost. No accounts, no cloud, no tracking.
Before installing VidClaw, you need OpenClaw already running on your server. If you haven't set that up yet, follow our complete OpenClaw setup guide first, then come back here.
You'll need:
One command does everything:
This installs Node.js and Git if missing, clones the VidClaw repo, installs dependencies, builds the frontend, and sets up a systemd service.
If you prefer to install manually:
The setup script handles everything: npm dependencies, Vite build, and systemd service registration.
Tip: If you prefer localhost-only access (e.g., you'll use an SSH tunnel), pass the --no-tailscale flag:curl -fsSL vidclaw.com/install.sh | bash -s -- --no-tailscale
VidClaw binds to localhost:3333 for security. To access it from your local machine, create an SSH tunnel:
Now open http://localhost:3333 in your browser. You should see the VidClaw dashboard with the Kanban board ready to go.
Tip: Add an alias to your shell config for quick access:alias vc="ssh -L 3333:localhost:3333 user@your-server"
Once you're in, you'll see six main panels:
Kanban Task Board — Your agent's work queue. Create cards, set priorities (low/medium/high/critical), and assign specific skills. Cards flow through four columns: Backlog, Todo, In Progress, and Done. Your agent checks this board automatically and picks up tasks ordered by priority.
Usage Tracking — Real-time token consumption and cost estimates. The progress bars match Anthropic's rate-limit windows, so you can plan heavy workloads without getting throttled.
Model Switching — Switch between Claude models (Sonnet, Opus, Haiku) directly from the navbar dropdown. VidClaw hot-reloads the OpenClaw config — no restart required.
Skills Manager — Browse all bundled and workspace skills. Toggle them on or off, or create custom skills. Changes take effect on the agent's next session.
Soul Editor — Edit SOUL.md, IDENTITY.md, USER.md, and AGENTS.md with a proper editor. Version history lets you revert if a personality tweak goes sideways. Six starter persona templates are included.
Activity Calendar — A monthly heatmap showing when your agent was active and what it worked on, parsed from memory files and task history.
With the dashboard open, give your agent something to do:
Your agent will pick it up on the next cron cycle (within 2 minutes) or heartbeat (within 30 minutes). Watch it move to "In Progress" and eventually "Done."
You can also use the quick-add bar at the top of the Kanban board. Type your task, use @skill to attach a skill, and hit Enter.
Head to the Soul Editor tab. Here you can shape how your agent thinks, writes, and behaves. The key files:
Pick one of the six starter templates to get going fast, then customize from there. Edit, save, and your agent picks up changes on the next session.
Note: You can switch models at any time from VidClaw's navbar — no need to edit config files manually once the dashboard is running.
Updating is a single command:
This pulls the latest changes, rebuilds the frontend, and restarts the systemd service.
Dashboard won't load?
./status.sh./logs.shlsof -i :3333Agent not picking up tasks?
openclaw gateway statusYou now have a fully working OpenClaw dashboard with VidClaw as your command center. Here are some things to explore:
Ready to take control of your AI agent?
VidClaw is free, open-source, and takes 60 seconds to install.
Get Started on GitHub →