YOUR INFRASTRUCTURE, AT A GLANCE
Environment overview
Servers, services, and the applications they power.
All environments · View filter only. Account permissions are unchanged.
Root filesystem usage measures the server disk. Docker categories come from Docker's storage accounting: shared image layers and cache can overlap, so the category sizes must not be added together. GB is decimal; GiB is binary.
WHAT NEEDS ATTENTION
Checking current observations…
Capacity at a glance
Latest collector snapshotsConnected servers
Snapshots every 30 seconds · stale after 2 minutesApplications
Container groupsECOSYSTEM · LIVE MAP
See how it all connects.
Follow a site from the web gateway to a server and the applications it runs. Select anything to explore it.
Swipe the map sideways to follow the connections →
Select an application to expand its containers. Bars show usage against known limits; striped tracks mean unknown limits or stale observations, never available capacity. Lines show configured relationships, not measured network traffic. Select a server for captured resource trends. Refreshes every 30 seconds; data is stale after 2 minutes.
| Container / image | Product / server | State | Health check | CPU | Memory | Disk · writable layer |
|---|
| Linux service | Server | State | Substate |
|---|
Loaded systemd service units, including inactive and failed units. Unloaded unit files are not included.
Dashboard access
Viewers can inspect inventory. Administrators manage accounts, routes, site access, captured logs, MCP access, server enrollment and reviewed maintenance actions.
| Username | Role | Status | Manage |
|---|
| When | Account | Action | Target |
|---|
Latest 100 account, enrollment, and cleanup events.
| Actions |
|---|
Hostnames created in Operations. Edit the destination, path, or gateway access through a validated preview. DNS is managed separately.
Site access users
These accounts protect website routes through a browser sign-in page and also accept Basic credentials. They are separate from dashboard and application accounts. Revoking an account or resetting its password ends its site access sessions on the next request. At least one account must remain.
Latest 20 previews and applies on this edge. Open a result for validation details.
| Requested | Administrator | Action | Result |
|---|
Server details & permissions
Account grants restrict environments, servers and operations on the server. The environment filter further narrows your permitted view. Route details omit credentials and sensitive header values.
HTTPS checks run from the collector host, approximately every 30 seconds. Two consecutive failures open an incident; two successes resolve it. These checks do not replace an external uptime monitor. Endpoint definitions live in the protected collector configuration.
Select an area on a server to see what needs attention and what remains unknown.
How to read these observations
Internet checks report connections from an independent probe to configured public addresses and ports. Local listeners alone do not prove reachability. Maintenance actions require an exact preview, impact review and recent identity verification.
Databases & mail
Inspect database pressure, storage and mail observations. Select a service to explore its captured history.
Can we recover?
Backup jobs, repository checks and restore drills are separate evidence. A green backup does not mean full application recovery has been proven.
Host maintenance
These are observed runtime settings. Firewall effectiveness, patch completeness and independent recovery still need separate verification.
Refreshes every 30 seconds. Up to 240 time buckets show averages, observed ranges and sample counts. Container histories are available from their detail panels; gaps remain visible. History starts when collection is enabled.
URGENT ALERTS, WHEREVER YOU’RE WORKING
Browser push notifications
Get a desktop popup when an application goes down, a server stops reporting, a container fails, or resource use reaches 95%. Your dashboard tab can be closed.
How delivery works
Push covers the servers and environments your account can access. Popups contain a general alert; open Operations for private details. Routine updates and log-watcher matches stay in the inbox. Alerts stop when you sign out or your sign-in expires. Use “Keep me signed in for 30 days” on your own device for longer coverage.
Allow notifications in both your browser and computer settings. Focus / Do Not Disturb can hide banners. Delivery also depends on your browser’s background settings and internet connection.
Incident inbox
The inbox follows the environment filter. Acknowledging an incident does not resolve it. All updates remain here, including recovery and log-watcher events.
Recent alert updates
Latest 300 matching lines. Logs may contain sensitive data despite best-effort redaction; access is limited to accounts with permitted log access and explicitly scoped AI credentials.
Log watchers
Global administrators can configure watcher scopes by environment, server, service and exact instance. Preview literal matches and review the threshold before saving. Recovery requires the selected quiet period with continuous complete capture; missing capture stays unknown.
Archiving & cleanup
Automatic cleanup runs every 30 seconds. Older captured logs move into gzip JSONL archives. Capacity limits take precedence: 20,000 searchable lines, 50 MiB / 500 archives; oldest archives expire first. Manual archiving moves up to 10,000 lines per click. These controls do not delete Docker’s original logs; configure Docker log rotation separately.
Connect your local AI assistant
Read-only MCP tools give Codex and Claude access to servers, container details, resource history, application health, incidents, and optionally captured logs.
Credentials are shown once, stored as hashes, and can be revoked here. Disabling the owning account revokes its access. The stdio bridge and setup instructions are in tools/ecosystem-ops/mcp_bridge.py and README.md. These tools cannot execute commands or modify servers.
Host worker credentials
Telemetry keys cannot change a server. These separate keys authorize only the chosen worker on one server and expire after 90 days. Install the shown value in that host’s protected collector configuration. Tokens are shown once.