OpenClaw Agent Dashboard: A Self-Hosted Monitoring Interface for OpenClaw

A Web Interface Purpose-Built for OpenClaw Agent Monitoring

OpenClaw Agent Dashboard solves a specific problem for developers running OpenClaw: there is no native web UI for inspecting what agents are doing, what API resources they are consuming, or how the host system is performing. The dashboard fills that gap with a browser-based interface that refreshes automatically every five seconds.

The target user is an individual developer or a small team that runs OpenClaw on a server or local machine and wants to check on agent sessions, review cost breakdowns by model and time period, browse memory files like MEMORY.md and HEARTBEAT.md, and manage running Docker containers, all from a single page rather than by reading log files or running CLI commands.

The server is written in pure Node.js with no npm runtime packages required. The optional dependencies (jq for Docker management, tmux and python3 for the Claude CLI usage scraper, and docker for the Docker management page) are documented separately.

What the Dashboard Tracks Across Seven Page Areas

The dashboard is organised around seven primary areas accessible via keyboard shortcuts 1 through 7.

The overview page shows session count, rate limit usage for Claude and Gemini APIs against rolling windows, lifetime stats (total tokens, messages, and cost since first session), and a 24-hour health history with sparklines for CPU, RAM, temperature, and disk. A provider toggle switches between Claude and Gemini usage, and a per-model selector lets the user choose which model or window to display (Opus, Sonnet, Pro, Flash, and others).

The sessions area displays all agent sessions with real-time activity status, session search and filtering by status, model, and date range, and a visual timeline of session activity. The cost page breaks down spending by model, session, and time period. The live feed streams agent messages across all active sessions in real time.

The memory viewer browses and reads the agent memory files stored in the workspace. The files manager views and edits workspace files, skills, and configurations, with additional security requiring re-authentication. The system page shows CPU, RAM, disk, temperature, and service health with quick-action buttons for restarting OpenClaw, the dashboard, or other services.

The security dashboard and config editor are gated behind a second authentication step and cover UFW firewall rules, open ports, fail2ban status, and SSH logs.

Installing and Starting the Dashboard

The quickest path is to clone the repository and start the server directly:

git clone https://github.com/tugcantopaloglu/openclaw-dashboard.git
cd openclaw-dashboard
node server.js

On first visit to http://localhost:7000, a registration screen appears where the user creates a username and password. The server prints a recovery token to the terminal on startup; saving this token is necessary to recover access if the password is forgotten.

To run the dashboard as a systemd service with auto-start and crash recovery, the repository includes an install script:

sudo ./install.sh

This creates /etc/systemd/system/agent-dashboard.service and an override configuration file, then enables and starts the service. To follow the service logs:

journalctl -u agent-dashboard -f

A Docker path is also available. The Dockerfile uses node:24-alpine as its base. The default container port is 3001, set via the DASHBOARD_PORT environment variable. The workspace directory is mounted from the host at the path passed to WORKSPACE_DIR.

Key environment variables include DASHBOARD_PORT (default 7000 in the direct launch path), WORKSPACE_DIR for the OpenClaw workspace path, OPENCLAW_DIR (defaults to $HOME/.openclaw), and OPENCLAW_AGENT (defaults to main).

Authentication Model: PBKDF2 Passwords, TOTP, and Session Security

Registration stores credentials using PBKDF2 password hashing. Login sessions can be set as session-only or as three-hour persistent logins via a remember-me option. The dashboard also supports TOTP multi-factor authentication compatible with Google Authenticator, making it suitable for deployment on a server accessible from a network.

Security headers include HSTS, a content security policy, rate limiting on login attempts, and timing-safe password comparisons to prevent timing attacks. Sensitive operations, including the security dashboard that shows UFW rules and SSH logs and the config editor that can restart the OpenClaw gateway, require re-authentication before the page is accessible.

An audit log event feed captures login events, configuration changes, and security events with an unread badge counter. The README notes that PROXY_SECRET can be set as a shared secret; if set, requests without the token in the path are rejected with a 403 response.

Limitations and Maintenance Status

The last push to the repository was on 2026-03-17. The repository is not archived.

The dashboard is specific to OpenClaw. Teams using Claude Code directly (without OpenClaw), or using a different agent runtime, cannot use this dashboard without modification. The session data, memory file paths, and agent configuration structures it reads are tied to OpenClaw's directory layout and conventions.

The repository metadata reports the license as unknown. The README displays an MIT license badge, but the license field detected by automated tooling is not resolved. Legal review is appropriate before deploying the dashboard in a commercial or regulated environment.

The most recent release, v3.0.0, was tagged on 2026-03-05. It added dark and light theme switching, the security dashboard, the config editor, Docker management, and browser notifications. There have been no releases after that date based on the repository data available.

For general server monitoring, Grafana paired with a metrics exporter is a well-known open-source alternative that covers CPU, RAM, disk, and service health across any infrastructure. The difference is that Grafana has no awareness of OpenClaw sessions, agent memory files, or Claude and Gemini API costs; it covers infrastructure metrics rather than AI agent workflow data.

Editorial conclusion

OpenClaw Agent Dashboard is the right tool for developers running OpenClaw who want a local web UI to track session activity, API costs, memory files, and system resources without writing custom tooling. It requires no npm runtime packages, installs with a single clone and node server.js, and can run as a systemd service. Developers who need remote access should review the Tailscale integration and the VPS deployment path described in the README. Verify the license terms before deploying in any commercial context, since the repository metadata reports the license as unknown despite the MIT badge in the README.

Frequently asked questions

Official sources

1. Issues
2. Project website
3. README
4. Releases
5. tugcantopaloglu/openclaw-dashboard on GitHub