Releases: caiopizzol/moor
v0.53.0
Now I'll generate the release notes based on my understanding of the changes.
What's New
-
Declarative container command and entrypoint — Set custom command/entrypoint on any project without a Dockerfile. Override the image default as an argv array, or omit to preserve the image default. Useful for stock images like cloudflare/cloudflared that need custom parameters.
-
File injection into containers — Declare files to write into a container before it starts, on every recreate. Content can be inline (for configs) or sourced from a project env var at create time (for secrets). Files land with the octal mode specified (e.g. 0600 for a TLS key).
-
New file management API — GET
/api/projects/:id/filesto list, POST to add/update (upserts by path), DELETE to remove files from a project. -
moor_deploy now supports command, entrypoint, and files — Pass them in a single operation. Files are additions/updates only — use
moor_file_removeto delete. -
New MCP tools —
moor_file_set,moor_file_list,moor_file_removefor declarative file management. Command and entrypoint parameters added tomoor_project_create,moor_project_update, andmoor_deploy.
How It Works
Files are written into the container via Docker's archive endpoint right before start, on every recreate. Command and entrypoint override the image defaults in the Docker create body — only set when provided, preserving the image default otherwise. Both features integrate seamlessly with existing build, redeploy, and cron workflows.
mcp-v0.27.0
@moor-sh/mcp-v0.27.0 (2026-06-12)
Features
v0.52.1
Fixes
- Fixed duplicate disk listings when bind-mounting a monitored path. The labeled monitored entry now takes precedence over the unlabeled container-visible duplicate.
v0.52.0
What's New
- Monitored host disks — Configure
MOOR_MONITORED_DISKSto bind-mount and report host filesystem usage alongside container-visible disks; appears in web UI, CLI stats, and MCP tool with optional friendly labels.
Improvements
- Disk monitoring now validates paths to prevent option injection (paths must be absolute and are passed with
--todf). - Unmounted configured paths are skipped instead of reported as empty.
- Updated
moor_statsMCP tool description to clarify it reports both container-visible filesystems and configured monitored disks.
v0.51.0
Based on my review of the code changes, here are the release notes for v0.51.0:
What's New
- All filesystems visible — Server stats now report every mounted disk, not just root. Catches hidden volumes (e.g. a 344 GB database disk at 49% while root looked empty). Falls back to root if
df --outputis unavailable. - Stats tab — New per-project tab shows live CPU, memory, network I/O, block I/O, and process count. Polled every 4 seconds while the tab is open.
Improvements
- Web, CLI, and MCP all surface the new disk and stats data.
- Backward compatible — the old single-disk
diskfield remains in/api/server/stats.
v0.50.0
What's New
- Project history — View CPU, memory, and network metrics over the last hour, day, or week. Web UI includes lightweight sparklines and a timeline of lifecycle events. CLI command:
moor history <project> [--hours N].
Fixes
- History argument parsing —
moor historynow accepts project name and--hoursflag in any order. Previously--hours 1 project-namewould silently misparse the flag value as the project ID.
v0.49.1
Fixes
- Container orphan cleanup — When container creation succeeded but start or network attachment failed (e.g., port conflict), the container was left in Docker without a database reference, preventing project deletion from cleaning it up. Containers are now force-removed on start failure, with the original error always surfaced.
- Cleanup error reporting —
removeContainernow throws on non-404 failures instead of silently logging, so cleanup logs accurately reflect whether orphan removal actually succeeded.
v0.49.0
What's New
- Project history — Query container resource usage and lifecycle events over time with the
moor_project_historyMCP tool or the/api/projects/:id/stats/historyAPI route. - Resource sampling — Moor now records CPU, memory, network, and block I/O metrics every 60 seconds, with automatic 30-day retention.
- Docker event stream — Captures container lifecycle events (start, die, oom, kill, stop, restart) directly from Docker, so short-lived state changes are recorded rather than inferred from polling intervals.
- Container correlation labels — Every moor-created container now carries
sh.moor.project_idandsh.moor.project_namelabels for consistent event attribution.
Fixes
- moor_rebuild error reporting — The tool now surfaces actual error messages when
/runfails before streaming output, instead of silently reporting "Rebuild complete" on validation errors or credential issues.