Running an AI agent on your own server is powerful, but managing it through config files and terminal commands gets old fast. VidClaw gives you a visual command center for your OpenClaw agent — everything from task management to cost tracking in one self-hosted dashboard. Here are the seven features that make it work.
The Kanban board is where you control what your agent works on. Tasks flow through three columns: Todo, In Progress, and Done. You create tasks, set priorities (P1 through P4), and your agent picks them up automatically in priority order.
Drag-and-drop lets you reprioritize on the fly. The board uses optimistic updates — your changes appear instantly while syncing to the server in the background. If something goes wrong, it gracefully falls back to a full refetch.
There's also a quick-add bar at the top. Type a task, use @skill to attach a skill inline, and press Enter. It's the fastest way to queue work for your agent without touching the mouse.
Tip: Use P1 for urgent work and P4 for low-priority background tasks. Your agent always picks the highest-priority item first.
AI API calls cost money, and it's easy to lose track. VidClaw's usage widget shows you exactly what your agent is spending — token counts, cost estimates, and usage trends — right in the dashboard header.
The widget pulls data from the OpenClaw API and updates on a 5-minute interval. You get a clear breakdown of input tokens, output tokens, and total cost so there are no surprises on your API bill.
Pair this with the model switcher in the navbar and you can instantly swap between models (like switching from Opus to Haiku for lighter tasks) to keep costs under control.
The calendar view gives you a bird's-eye view of your agent's activity over time. Each day shows what tasks were completed, what actions were taken, and how active your agent was.
This is particularly useful for spotting patterns — maybe your agent is most productive on certain days, or maybe a particular type of task consistently takes longer than expected. The activity log keeps the last 200 entries, enough to see meaningful trends without bloating your storage.
Skills are reusable capabilities you can attach to tasks. Think of them as specialized instructions — a "code-review" skill might tell your agent to check for security issues, while a "blog-writer" skill might define a particular writing style and format.
The Skills Manager lets you browse, enable, disable, and create skills from the UI. Each skill is a simple text definition that shapes how your agent approaches a specific type of work. You can attach skills to individual tasks on the Kanban board using the @skill syntax.
The Soul Editor is where you define who your agent is. It's the system prompt — the core instructions that shape your agent's personality, communication style, and default behavior across all tasks.
VidClaw ships with starter templates to get you going, or you can write your own from scratch. Want a concise, no-nonsense agent? A friendly collaborator? A cautious code reviewer? The soul is where you make that happen.
Changes take effect immediately — no restarts needed. Your agent picks up the new soul on its next task.
Your agent works inside a workspace directory on your server. The File Browser gives you a visual interface to explore those files without SSHing in and navigating manually.
Browse directories, view file contents, and see what your agent has been creating or modifying. It's especially handy for reviewing generated code, checking config files, or verifying that a task produced the expected output.
VidClaw doesn't poll for changes — it uses WebSocket connections to push updates to your browser the moment something happens. When your agent picks up a task, finishes a job, or logs an activity, the dashboard reflects it instantly.
This means you can keep VidClaw open in a tab and always see the current state of your agent without refreshing. The Kanban board, heartbeat timer, and activity feed all update in real time.
How it works: The server broadcasts state changes after every mutation. The useSocket hook on the frontend subscribes to specific message types, so each component only re-renders when its data changes.
All seven of these features run on your server, accessed through an SSH tunnel. VidClaw binds to localhost only — it's never exposed to the internet. There are no accounts, no cloud dependencies, no analytics tracking your agent's data. You own everything.
This is what makes self-hosted AI management different. Your tasks, your agent's memory, your API keys — they all stay on hardware you control.
VidClaw installs in under a minute on any Linux VPS with Node.js. If you already have OpenClaw running, it's a single command:
curl -fsSL vidclaw.com/install.sh | bash
Check out the full setup guide for step-by-step instructions, or head straight to the GitHub repo to get started.
Your agent deserves a proper dashboard.
VidClaw is free, open-source, and self-hosted. Give it a try.
View on GitHub →