Writes use only <ID>_ADMIN_TOKEN and need the apps.write capability.
Store URLs are resolved from each server's own listing. The config is saved
to history before an uninstall. Zip uploads are fanned out to the servers.
Every target's result is audited. The console gets a row menu on the
matrix, a target picker, a results dialog with Retry failed, the store and
install from file in All servers mode, and single-server actions for writers.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
hub/apps/matrix reads apps/list and each app's config on every server (the read
token first, the admin token once if a server refuses) and returns per app and server
the version, available update and a config hash, plus whether versions and configs
match and how many JSON leaf paths differ. Config texts never leave the hub. In hub
mode the Apps section has the server selector: All servers shows the matrix with
filters for differing versions, differing configs and updates; one server shows its
list without write actions. The dev mock takes --variant a|b with differing apps.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
node:http server that serves the console build and a /api/* the console accepts:
its own login and session (clusterInitialized, the servers as nodes, hub flag,
Dashboard and Logs only), node=<id> forwarded with that server's token, no node or
cluster merged (stats, top lists, query logs, apps), every write refused with 403,
a short response cache, and an unreachable server reported as hubWarnings. The dev
mock takes --port and --scale so two of them differ. Vitest skips .claude/ worktrees.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>