A dashboard for OpenClaw that solves the 'where is localhost' puzzle

What it does OpenClaw Studio is a web dashboard for the OpenClaw Gateway. You point it at a gateway—local, remote, or cloud—then manage agents, chat, job approvals, and configuration from one place. It is a companion interface rather than a standalone runtime.

The interesting bit The architecture deliberately splits traffic: your browser talks to Studio over HTTP and SSE, while a single server-owned WebSocket connects Studio to the gateway. Studio also keeps a local SQLite database (runtime.db) to project runtime state and replay SSE history, so a browser refresh does not reset your context.

Key highlights

• Server-side control plane: the Node process owns the gateway connection and persists state in SQLite, not the browser.
• Explicit “localhost means the Studio host” mental model; the documentation invests heavily in preventing remote-access topology confusion.
• Supports local, split, and cloud-cloud deployment topologies; Tailscale is recommended for secure remote access.
• Requires STUDIO_ACCESS_TOKEN when bound beyond loopback, enforcing access control for exposed instances.
• Ships with documented UI workflows, permission sandboxing rules, and a semantic color system to keep behavior consistent.

Caveats

• The README assumes you already know what OpenClaw is and does not define gateway capabilities or what “agents” do.
• Native module (better_sqlite3) version mismatches can block startup; repair scripts exist, but it is a documented friction point.
• Serving Studio under a subpath like /studio requires rebuilding with a configured basePath.

Verdict Worth a look if you run an OpenClaw Gateway and want a clean, persistent web UI for managing agents and approvals. Skip it if you are not already running OpenClaw; this is strictly a companion dashboard, not a standalone product.