CoreScope 0.7.0

Added

• Machines › why a systemd unit failed: the units in a Failed systemd units alert are
  clickable. CoreScope runs systemctl status <unit> --no-pager -n 0 and the unit's journal over
  your BatchMode ssh (read-only, no sudo, unit name validated), groups the last run's
  error/failed lines with counts and shows a hint for well-known causes — logrotate
  unknown user (a config left for a deleted site user, e.g. Laravel Forge's
  /etc/logrotate.d/forge-*), logrotate duplicate log entry, postrotate failures and nginx -t
  errors. Says when journald hides system logs from the ssh user (not in adm/systemd-journal).
  Hints only explain; CoreScope never changes the server.
• MCP explain_failed_unit {host, unit}: the same status, grouped errors and hints for agents.
• Storage: processes using the disk now; SSD health from NVMe SMART (wear, spare, data written
  and read, power-on hours, cycles, unsafe shutdowns, media errors, verdict); What grew today
  opens each folder to the subfolders that grew and links to the cleanup plan; a safe cleanup
  plan for caches tools re-create, with estimates, the exact command or folders and busy-tool
  checks; large files in the home folder and its linked drives (one filesystem per root,
  cloud storage and node_modules skipped, app-managed files only shown); Eject… shows the
  processes holding a volume and runs diskutil eject. CLI top --disk, disk-health,
  space --plan, large-files, eject-check; MCP disk_health, cleanup_plan,
  run_cleanup_plan, large_files, volume_holders, eject_volume.
• Databases: Open in TablePlus, Redis Insight, Sequel Ace, Postico or DBeaver and Copy DSN;
  Redis keys by prefix (sampled, adaptive depth) with keys without TTL and FLUSHDB after a
  confirmation; snapshots of MySQL/MariaDB and PostgreSQL databases (inside a container with
  its own credentials, never seen by CoreScope) and restore. CLI db snapshot|snapshots|restore| redis; MCP db_snapshot, db_snapshots, db_restore, redis_keyspace, redis_flush; the MCP
  instructions ask agents to snapshot before migrate:fresh or db:wipe.
• Claude limits without the status line: from Claude Desktop's own samples
  (plan-usage-history.json), local and credential-free. Pace per limit window (projected at
  the reset, or when it reaches 100%). MCP usage_limits.
• Settings backup outside ~/Library (default ~/.corescope/backups, or a folder you pick),
  written a few seconds after a change; a launch with empty settings offers to restore it. MCP
  backup_settings.

Fixed

• Status line data that carries the reset epoch in used_percentage is ignored instead of shown.
• Sizes and rates read "0 KB" instead of "Zero KB".

Commits since v0.6.0

• feat(machines): explain why a failed systemd unit failed (ff80a5a)
• feat(storage): SSD health, who's writing, growth drill-down, cleanup plan, large files, eject (17fddd3)
• feat(databases): open in TablePlus/Redis Insight, Redis by prefix, snapshots (59a64da)
• feat(agents): Claude limits from Claude Desktop, pace, MCP usage_limits (c1dff46)
• feat(settings): automatic settings backup outside ~/Library, restore on launch (4bde9b0)
• chore(release): 0.7.0 (c01e9b2)

Full diff: v0.6.0...v0.7.0

Install

1. Download CoreScope-0.7.0-macos-universal.dmg below (universal: Apple Silicon and Intel, macOS 14+).
2. Open the DMG and drag CoreScope to Applications.
3. The app is open source and signed ad hoc, not notarized by Apple, so macOS blocks it the first time. Open it once, then go to System Settings → Privacy & Security and click Open Anyway. Alternatively, in Terminal:

   xattr -dr com.apple.quarantine /Applications/CoreScope.app

The Intel slice is built and checked on every release but has not yet been run on an Intel Mac; reports are welcome.

Checksums (SHA-256)

e3219ef26580703c7c8e866c1dc030247069d62568d9602b74aa1ff0a1707535  CoreScope-0.7.0-macos-universal.dmg
76f0b21608f60889416ab171a8f386cf09c7c947d6e566166a7ea2b1bb34cda0  CoreScope-0.7.0-macos-universal.zip

Verify with shasum -a 256 -c SHA256SUMS.txt in the download folder.