Command Center runs standalone. Spacesuit is its highly recommended companion for agent flavors and domain workflows. Spacesuit scaffolding and collectors can run on their own, but Spacesuit does not provide a graphical human–agent interaction surface: use Command Center for the dashboard and supported controls. Neither package adds a new chat composer through this extension API.
| Selection | Behavior |
|---|---|
COMMAND_CENTER_MODE=core |
Core monitoring only; zero extension or legacy domain reads |
COMMAND_CENTER_MODE=legacy |
Existing built-in Intel/Pipeline/Monetization panels; zero extension reads |
Workspace selection with mode: "extensions" |
Validated snapshots; legacy collectors disabled |
COMMAND_CENTER_MODE=extensions |
Requires valid workspace selection; fails closed without it |
| No environment override and no selection file | Legacy compatibility mode; Spacesuit not required |
| Invalid mode/selection or wrong identity | Core monitoring plus a safe diagnostic; no domain fallback |
This first release preserves existing built-ins by default rather than silently
removing them. A fresh install already works without Spacesuit; choose explicit
core mode for a minimal dashboard. Full removal of the compatibility modules is
future migration work, not part of v1.
# Run from the Command Center checkout, with normal OpenClaw workspace discovery:
COMMAND_CENTER_MODE=core npm startThe optional selection file is state/command-center/extensions.json, under the
selected workspace. Environment mode takes precedence. Profile is OPENCLAW_PROFILE
(or empty), agent binding is OPENCLAW_AGENT (or main). Generic monitoring still
uses this release's existing main-agent session adapter; setting OPENCLAW_AGENT
binds extensions, not a newly implemented arbitrary-agent monitoring feature.
Use a separate workspace for each profile/agent selection; sharing this selection
file across different identities is deliberately rejected.
{
"schemaVersion": 1,
"profile": "",
"agentId": "main",
"mode": "extensions",
"enabled": [{ "id": "spacesuit.intel", "version": "1.0.0" }]
}IDs match ^[a-z][a-z0-9-]*\.[a-z][a-z0-9-]*$. Version pins must match exactly.
Duplicate IDs, unknown schema versions and identity mismatches fail closed. No
manifest may specify an executable, route, URL, or data path for the host to load.
Each selected snapshot lives at
state/command-center/extensions/<id>.json. Spacesuit's reviewed collector writes
it atomically; Command Center never runs collectors in a request or schedules them.
{
"schemaVersion": 1,
"id": "spacesuit.intel",
"version": "1.0.0",
"profile": "",
"agentId": "main",
"observedAt": "2026-01-01T00:00:00Z",
"status": "ready",
"panels": [
{
"id": "summary",
"type": "metrics",
"title": { "en": "Research", "zh-CN": "研究" },
"metrics": [{ "label": { "en": "Files", "zh-CN": "文件" }, "value": 3 }]
}
]
}- Producer status:
ready,unavailable, orerror. Host derivesstaleafter five minutes. Missing files become unavailable; invalid files become error. metricscontains localized label/value pairs; values are finite numbers or strings.tablecontainscolumns: [{key, label}]androws: [{<key>: value}]. Every declared cell must be a finite number, string or explicit null.- All titles/labels carry nonempty
enandzh-CNvalues. Panel/column IDs are unique safe keys. The host strips unrecognized fields rather than exposing them. - v1 uses namespaced extension/panel IDs and bilingual labels embedded in each snapshot. No locale-file fetching or translation registration is needed at runtime.
- Data values are rendered as text, never HTML, links or executable attributes. Business values are excluded from heuristic phrase translation. Both labels and status messages switch language while the panel is open.
Selections: 16 KiB and 8 extensions. Each snapshot: 256 KiB, 8 panels, 24 metrics, 12 columns, 100 rows/table, 2,048 characters/value and 256 characters/label. Oversized or invalid snapshots are rejected, not silently rendered in part. Symlinks, nonregular files and paths outside the workspace are rejected.
Reads refresh asynchronously every five seconds on access and are coalesced.
Selection and individual snapshot reads have bounded waits; a failed or stalled
snapshot does not suppress healthy siblings. Pending disk reads are capped so a
stalled filesystem cannot accumulate work. /api/state returns cached normalized
extensions: {schemaVersion, mode, items, diagnostic?} through the existing state
and live-update paths. There are no extension-specific browser requests or streams.
Core monitoring does not await snapshot refreshes. Diagnostics contain fixed codes, not raw exceptions or private paths. Missing/stale data never becomes a false zero. Core/legacy explicit modes perform no selection/snapshot reads at all.
Only enable reviewed producers and approved data sources. Snapshots are display-ready data, not a place for credentials. Host and companion collectors apply common credential-pattern redaction as defense in depth; this is not a comprehensive secret detector or a substitute for selecting appropriate data. Snapshot contents are visible to users authorized to access the dashboard. Existing hostname/session privacy toggles do not hide arbitrary business fields in extension tables.
The existing dashboard authentication boundary applies. This API adds no permission or public access. Installing a package or selecting a flavor does not authorize network calls, writes to other profiles, scheduled collection or deployment.
Use the Spacesuit flavor CLI to preview, apply and collect. See the companion README for exact commands. Keep COMMAND_CENTER_MODE unset when using workspace selection. Applying a flavor changes only that workspace's selection; it does not start Command Center or collectors. Collection is an explicit separate operation. Disabling selects core mode while retaining data/snapshots; restore recovers the previous selection from a backup. Review package versions together before upgrading.