What shipped.

Every release of moin, newest first. Follow it with RSS, in your reader or a Slack channel.

v1.0.0-beta.25

November 5, 2026

In short: a new terminal UI, moin login through the browser and a new home page: the fastest way to run your apps, from any screen.

New

  • A new terminal UI: moin opens a dashboard of your apps and servers, space shows everything you can do, space space searches apps, servers and actions, esc goes back.
  • Instant start: the terminal UI shows your last view at once and refreshes behind it; switches and env changes show before the server answers.
  • Log in through the browser: moin login and the terminal UI show a code, you approve it in the browser with GitHub, an email link or your password, done. No token to copy.
  • Your computers under Tokens: each browser login shows as “CLI on ”, ends after 90 days without use, and you can sign it out one by one.
  • Where things work: a docs page lists every feature for web, CLI, terminal UI, API and MCP, with the reason where one is missing.
  • A new home page: one calm screen with a wall of app containers, then everything moin does in tabs, live in two minutes, security, backups, four ways in and a comparison.
  • Built in the open: the home page shows how often moin ships, read from the changelog.

Changed

  • One promise everywhere: Deploy anything. Lose nothing. Your apps on your servers. Every change backed up. Every restore tested. Home, sign-in, docs, CLI and catalog say the same.
  • Deploy now: the main button says Deploy now everywhere, instead of Get started or Start free.
  • Calmer navbar: on a phone only the logo and the menu; Apps has no New badge; GitHub stars show from 1,000 on.
  • Comparison in black and white: color only means status now.
  • moin login in a terminal: it always goes through the browser; pass a token with --token, on stdin or as MOIN_TOKEN.
  • Setup in the web: the terminal UI covers daily work; team, billing, adding servers, GitHub, the catalog and SQL open the right page in the browser.

Fixed

  • http on your tailnet: the CLI connects over plain http to short tailnet names like your server’s name, not only to IP addresses.

Upgrade

Self-hosted: Update now, or sh update.sh. CLI: curl -fsSL https://moin.run/cli.sh | sh.

v1.0.0-beta.24

November 5, 2026

In short: the old name ops is gone: only moin and its names work now, so update scripts, CI and compose files that still say ops.

Changed

  • Guides: a small book icon on each card opens its guide, instead of a “How it works” link.
  • x-ops stops a deploy: a compose.yml with an x-ops block fails with a message to rename it to x-moin, instead of counting.
  • Instant address user: the user in front of a protected instant address is moin, and moin protect asks for moin by default; it takes effect with the app’s next deploy.
  • Rules file: the agent reads only /etc/moin/rules. Move /etc/ops/rules there if you still have one.

Fixed

  • Switches stay where you put them: every switch saves only its own setting and shows what was saved, so “Deploy every push” no longer flips back when GitHub is slow.
  • Uptime pause sticks: pausing a check holds, even when a check was running at that moment or the control plane restarts.

Removed

  • The ops command: use moin. An agent still installed as ops no longer updates itself; install moin on that server again.
  • OPS_* variables: use MOIN_URL, MOIN_TOKEN, MOIN_STATE_DIR, MOIN_RULES, MOIN_IMAGE and MOIN_PULL.
  • The old CLI config folder: run moin login once if the CLI says you are not logged in.
  • Old downloads and image: /download/ops-* and ghcr.io/axadrn/ops-controlplane get no new releases; use /download/moin-* and ghcr.io/axadrn/moin.
  • X-Ops-Signature: webhooks carry only X-Moin-Signature.
  • The move from ops: update.sh no longer copies /etc/ops-controlplane, and settings that point to /run/ops-controlplane need /run/moin-controlplane.

Upgrade

Self-hosted: Update now, or sh update.sh; agents update themselves. CLI: curl -fsSL https://moin.run/cli.sh | sh.

v1.0.0-beta.23

November 4, 2026

In short: page counts only on the public site, never inside your workspace.

Changed

  • Analytics only where it belongs: the home page, docs, changelog, catalog and legal pages are counted; the app, the admin and sign-in never are, so no workspace or app name and no sign-in link reaches a page counter.

Upgrade

Self-hosted: Update now, or sh update.sh.

v1.0.0-beta.22

November 4, 2026

In short: every app in the catalog speaks moin now, so installs keep working after the old name goes on December 4.

Fixed

  • Catalog after December 4: the 47 apps still used the old x-ops block in their compose files and would have stopped installing then. They use x-moin now.
  • Messages: the last hints that still said ops deploy say moin deploy.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves.

v1.0.0-beta.21

November 4, 2026

In short: the site says everything moin does today: Postgres next to SQLite everywhere, a fair comparison with sources, and answers to the questions people ask now.

Changed

  • Postgres everywhere it belongs: the homepage, the comparison and the docs name Postgres next to SQLite, for backups to any second, the daily restore check and the data browser.
  • A fairer comparison: a new row for Postgres backups, every competitor fact from their own docs.
  • Answers that are true today: any language deploys without a Dockerfile, the catalog, instant addresses, moin exec, and alerts for backups, deploys and full disks.
  • Roadmap: what shipped moved to Now; MySQL backups are next.
  • A new footer: product, docs, open source and company, at a glance.
  • Why moin: its own section on the homepage, linked under the buttons.

Fixed

  • Self-hosting commands: the install URL and git clone pointed at the old repository name.

Upgrade

Self-hosted: Update now, or sh update.sh.

v1.0.0-beta.20

November 4, 2026

In short: ops is now moin, at moin.run. Deploy anything. Lose nothing. Everything you run moves over by itself, and nothing breaks on the way.

Why

moin is how Hamburg says hello. Hamburg is a container port, and your apps run in containers too. The story.

Changed

  • New name, new home: moin at moin.run, the command moin, the logo moin.. The old address sends you to the new one.
  • Your servers move by themselves: the next agent update renames everything on the server to moin, copies each app’s data and checks it file by file. Apps keep serving while it moves.
  • Backups go on: Postgres starts a fresh continuous backup under its new name right away. Every older snapshot and point in time still restores.
  • Self-hosted moves too: update.sh copies your settings and data to the new folders and starts moin from the image ghcr.io/axadrn/moin. The old folders stay.
  • Old names keep working until December 4, 2026: the ops command, x-ops in compose.yml, OPS_* variables, the old download paths and image, and the X-Ops-Signature webhook header next to the new X-Moin-Signature.
  • One address for everything: write to [email protected].

Upgrade

Hosted: nothing to do. Self-hosted: Update now, or sh update.sh; agents follow by themselves. CLI: curl -fsSL https://moin.run/cli.sh | sh.

v1.0.0-beta.19

November 4, 2026

In short: 47 apps in the catalog, with AI on your own server: Open WebUI, Ollama, Flowise, plus Metabase, Outline, Twenty, Directus, Jellyfin and more.

New

  • 17 more apps: Open WebUI, Ollama, Flowise, Activepieces, Metabase, Outline, Twenty, Directus, Appsmith, Jellyfin, Audiobookshelf, Trilium, Homepage, Gatus, Healthchecks, PrivateBin and Seafile. Each one installed for real before it got in.
  • An AI category: chat with your own models, next to your data.

Fixed

  • Newest releases from GitHub’s registry: ops read only the first page of tags, so some apps showed an old release as the newest. It reads them all now.

Upgrade

Self-hosted: Update now, or sh update.sh. The catalog comes with it.

v1.0.0-beta.18

November 4, 2026

In short: run a command in your app’s container from anywhere, no SSH: migrations, the first admin, a quick look.

New

  • Run a command: ops exec shop -- rails db:migrate, or --service db -- psql -c "…". The output streams live and the command’s exit code comes back. Web: Run a command on the app page. Also the TUI :exec, the API and MCP exec_command.
  • Safe by default: only owners and admins, never through a shell, at most an hour, every run in the activity with its exit code. exec: off in /etc/ops/rules turns it off for a server.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves; an older agent answers that it needs the update first.

v1.0.0-beta.17

November 4, 2026

In short: every app is online at once, with HTTPS and a password, before it has a domain; and each service of an app can have its own.

New

  • Instant address: after its first deploy every app answers at <app>.<your-server-ip>.sslip.io with HTTPS, guarded by a password until you open it. Switch to nip.io if one is ever down. Web: Instant address on the app page. CLI: ops instant, ops instant nip.io, ops instant open.
  • Domains per service: shop.com for web and api.shop.com for api in one app. In compose.yml under x-ops.domains, or ops domains add --service api api.shop.com.
  • Uptime for every service: each domain of each service is checked every minute.

Changed

  • ops deploy ends with the address the app is live at.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves; servers behind a Cloudflare Tunnel get no instant address, since it needs ports 80 and 443 open.

v1.0.0-beta.16

November 4, 2026

In short: one page tells you whether everything is OK, and ⌘K takes you anywhere.

New

  • Overview: the first page of a workspace. “All good.” when it is, otherwise what needs you: a server offline, a failed deploy, a backup that stopped, a domain down, a full disk. Each with the way to fix it.
  • ops status: the same in your terminal, and it exits 1 when something needs you, for scripts and CI. Also the TUI :status, the API and MCP get_status.
  • ⌘K: jump to any app, server, bucket or page, and deploy, restart or roll back from the keyboard. ? shows every shortcut.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves.

v1.0.0-beta.15

November 4, 2026

In short: 30 open-source apps on your server in one click, any version, updates with a backup first: Umami, n8n, Grafana, Nextcloud, Vaultwarden, WordPress and more.

New

  • App catalog: browse at /apps, install from the Apps page, ops catalog install umami --server web-1, the TUI, the API or MCP. ops makes the secrets, sets HTTPS on your domain and backs up its Postgres or SQLite.
  • Every version: pick any release, newest first with its date, fetched from Docker Hub and GitHub once an hour.
  • Updates: “Umami 3.5.0 is out” on the app page, ops update <app>, with a snapshot first. An update.available alert tells you once per release.
  • RSS for the changelog: /changelog.xml, and a dot in the account menu when there is something new.

Changed

  • A calmer navbar: Pricing, Docs, Changelog, and Apps on its own. Roadmap moved to the footer, and Get started is the one way in.
  • The app sidebar: no app list and no shortcut letters; Docs, Support and the changelog are in the account menu.
  • Health checks: a redirect counts as up, and a slow first start gets two minutes.

Fixed

  • Docs on a phone: a long page title no longer runs under the search.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves.

v1.0.0-beta.14

November 4, 2026

In short: self-host ops on one server with one command, and the home page shows in one table how fresh your backups are.

New

  • One server, one command: curl -fsSL https://raw.githubusercontent.com/axadrn/ops/main/controlplane/deploy/install.sh | sh -s -- --domain ops.example.com --email [email protected] installs Docker, the control plane and the agent, and serves it on your domain with a Let’s Encrypt certificate. It prints your sign-in link.
  • Three ways to run ops: hosted, one server, or two servers. The self-hosting guide says when to pick which.

Changed

  • Self-hosted without email or a bucket: a self-hosted control plane starts without them. Sign-in links go to its log, and docker compose -f /opt/ops/compose.yml exec controlplane ./main sign-in-link prints one. Add email and a bucket any time.

Fixed

  • Apps that wait for their database: a service that restarts once while its database starts no longer fails the deploy. ops now waits until it has run for 10 seconds without a restart, up to 30 seconds.

Upgrade

Self-hosted: Update now, or sh update.sh. Nothing changes for a control plane on its own server.

v1.0.0-beta.13

November 4, 2026

In short: you hear it when a backup stops, a restore check fails or a disk fills up, and servers clean up old images by themselves.

New

  • Backup alerts: a backup that stops working, or a daily restore check that fails, sends an alert by email, Slack, Discord, Telegram or webhook, once, and again when it works. ops alerts add …, events backup.failed, drill.failed.
  • Disk alerts: a server over 90% full sends an alert, and another when it is back under 85%. Event disk.full.
  • Disk cleanup: after every deploy and once a day, servers remove images no release needs anymore, dangling images and build cache older than a week. Rollback images and volumes stay.

Fixed

  • Restore check while a database is down: it no longer fails just because the Postgres service is stopped.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves. Destinations that get every event also get the new ones.

v1.0.0-beta.12

November 4, 2026

In short: deploy any app without a Dockerfile: ops reads your repo and builds Node, Bun, Python, Ruby, PHP and more.

New

  • No Dockerfile needed: a repo without compose.yml, Dockerfile or Go module now builds with Railpack. ops finds the language, its version and the start command, and runs it on PORT 8080. The deploy log says what it found, such as Railpack: Node 22, start: npm run start.
  • Build secrets stay secret: your env vars reach the build as secrets, never in the image.
  • Start command of your own: set RAILPACK_START_CMD with ops env set.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves.

v1.0.0-beta.11

November 4, 2026

In short: ${VAR} in compose.yml takes your env vars from ops, so a Postgres service gets its password; update if your compose uses variables.

Fixed

  • Variables in compose.yml: ${DB_PASSWORD} and the like now take the app’s env vars from ops, as a .env next to the file would. Before, they came out empty and a service such as Postgres did not start.
  • Deploys while a database is down: a Postgres that does not run no longer blocks the deploy that could fix it. The deploy log says there was no snapshot, and why.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves; then deploy the app again.

v1.0.0-beta.10

November 4, 2026

In short: Postgres gets what SQLite has: every change backed up, full copies, a daily restore check and the data browser. And a new home page: Deploy anything. Lose nothing.

New

  • Postgres backups: every change goes to your bucket within a minute, seconds when it’s busy, with WAL-G. Go back to any second of the last 7 days: ops restore --at "2026-11-04 14:02".
  • Postgres snapshots: a full copy every hour and before every deploy, restore and edit. ops backup, ops restore <snapshot>.
  • Checked every day: ops restores the newest Postgres copy into a throwaway database and checks it, like SQLite.
  • Postgres in the data browser: tables, rows, schema, search, edit and delete by primary key, and read-only SQL. Web: the Data tab. CLI: ops db, ops sql.
  • Rows by primary key: SQLite tables without a rowid can be edited too.

Changed

  • Home page: “Deploy anything. Lose nothing. Open source.” The product shots change by themselves.
  • Reads stay reads: SQL that changes data is refused up front in a read, with a hint to use --write.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves; deploy an app once so ops finds its Postgres.

v1.0.0-beta.9

November 4, 2026

In short: a new domain on a server without a Cloudflare Tunnel gets its HTTPS certificate again; update if you add domains.

Fixed

  • HTTPS for new domains: since beta.1, a domain added on a server that serves ports 80 and 443 itself got no Let’s Encrypt certificate and showed a browser warning. It gets one again on the next deploy. Servers behind a Cloudflare Tunnel were never affected.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves; then deploy each app with a new domain once.

v1.0.0-beta.8

November 4, 2026

In short: the app page puts Deploy from GitHub first; deploying from CI waits behind one click.

Changed

  • Deploy from CI: deploying an image and the deploy hook now sit folded under Deploy from CI on the app page, so the GitHub Deploy is the one you see. An app with a hook shows it open.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves.

v1.0.0-beta.7

November 4, 2026

In short: apps and servers are slim cards you drag as a whole, the version of ops is always in sight, and changing repos on github.com no longer ends on an error.

New

  • Version in the sidebar: the control plane’s version sits at the bottom of the sidebar and opens the changelog.
  • Update pending: a server whose agent is behind says so next to its version; it updates itself on its next connect.

Changed

  • Apps and servers lists: every app and server is a slim card of its own, with its grip on the left, the same on both pages.
  • List and grid switch: icons only.

Fixed

  • GitHub repos changed on github.com: coming back to ops no longer shows “GitHub is not connected”; the change was saved all along.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves.

v1.0.0-beta.6

November 4, 2026

In short: every agent now updates itself, so metrics arrive everywhere, charts move to a file of their own to keep the database small, and the app tabs fit on a phone.

Changed

  • Charts in their own file: metrics and hourly uptime live in history.db next to the database, without a backup. The database stays small and a restore fast. The charts start over once; outages and uptime state stay.
  • Home page: the live numbers show from 1,000 deploys on.

Fixed

  • Agents that never updated: an agent built from a commit instead of a release, such as one installed from main, now updates itself like any older agent.
  • Metrics page: no longer tells you to update agents by hand; they update themselves.
  • App tabs on a phone: Overview, Data, Backups, Uptime and Metrics no longer push the page wider than the screen; swipe them.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves.

v1.0.0-beta.5

November 4, 2026

In short: alerts go to Slack, Discord, Telegram and webhooks, every server and app has live charts, and each app can back up to its own bucket.

New

  • Alerts where your team is: send alerts to Slack, Discord, Telegram or a signed webhook, as well as email, with a test for each. Web: Alerts. CLI: ops alerts add slack <name>. Also the TUI settings, API and MCP.
  • Live metrics: CPU, memory, disk and network of every server and app, live for the last hour and back 30 days, with each deploy marked. Web: the Metrics tab. CLI: ops metrics shop. TUI: m. API and MCP too.
  • More than one bucket: give each app its own bucket for uploads and backups, say one per client. Web: Buckets. CLI: ops buckets add, ops buckets use.

Changed

  • Buckets: the bucket you had is now the default one. Older backups stay where they were and restore from there when an app moves.
  • Request and error counts per app are not in the metrics yet; response times are on the Uptime tab.

Upgrade

Self-hosted: Update now, or sh update.sh. Agents update themselves; until then, metrics and per-app buckets wait for the new agent.

v1.0.0-beta.4

November 3, 2026

In short: ops watches your sites every minute and mails you when one goes down, puts apps and servers in your order, and updates itself and its agents.

New

  • Uptime checks, built in: every domain of every app is checked from outside once a minute, from the first deploy on, with nothing to set up. The app’s Uptime tab shows 24 hours, 7 and 30 days, the certificate and the last outages. Also ops uptime, the TUI (u on an app), API and MCP.
  • Your own order: drag apps and servers where you want them, or ops apps move blog top, K / J in the TUI. Everyone in the workspace sees the same order everywhere.
  • List or grid: switch how the Apps and Servers pages look; ops remembers it for you.
  • Update now: a self-hosted ops shows when a release is out, with its notes, and updates in one click. If the new version does not start, the old one comes back. Also ops update.
  • Agents update themselves to the control plane’s version, one server at a time and never during a deploy. Turn it off with ops servers auto-update off.
  • Email alerts: owners and admins get a mail when a site goes down or comes back, a server goes offline or a deploy fails. Turn it off for you with ops alerts off.

Changed

  • Activity: a server going offline and coming back now shows up there, and so does every site that goes down and comes back.
  • Self-hosted gets releases only: update.sh takes the newest release, not every build in between.
  • A copy of the database before every upgrade: ops writes backup-before-<version>.db next to it before it changes the schema.

Fixed

  • Signing in with GitHub finds your account by any verified email on your GitHub account, not only the primary one, instead of making a new empty account.

Upgrade

Self-hosted: git pull in the repo, then sh update.sh once, which also starts the updater. From then on: Update now.

v1.0.0-beta.3

November 3, 2026

In short: Vim keys in the terminal, env vars as one .env file, and a dead server can finally go.

New

  • Vim keys in the TUI: j k, gg G, / to search, : for commands, ? for help. Logs scroll, and G follows the end again.
  • Env vars as one .env text: paste or edit them all at once. Web: Fields | Text. CLI: ops env push. TUI: e, then edit. API PUT /api/v1/apps/{app}/env, MCP replace_env. ops shows what changes and asks first.

Changed

  • TUI keys moved to make room for vim: GitHub g → gh, logs l → L, rename n → cw, deploy hook h / H → :hook / :hook off.
  • Admin is only on the hosted control plane: self-hosted, you manage people under Team; ADMIN_EMAILS only says who may sign up.
  • Free forever is what you tick (hosted Admin): Admin shows each user’s plan (trial, trial ended, free forever, paying) and turns free forever on or off.

Fixed

  • A dead server can be removed: offline, it goes with its apps, which stay on the machine if it comes back. Online, its apps still go first.

Removed

  • The waitlist: sign-up is open. The first start deletes the waitlist table with its addresses; whoever is in keeps their plan.

Upgrade

Self-hosted: pull the new image and restart. Agents need no update.

v1.0.0-beta.2

November 2, 2026

In short: Sign in with GitHub in one click, and the first page asks only your name and your workspace.

New

  • Sign in with GitHub first: “Continue with GitHub” is the main way in; the email link stays below it, and a password is optional in the settings.
  • One short first page: your name and your workspace name. From GitHub both are filled in, so it’s one click. Invited people only give their name.
  • API: POST /api/v1/auth/onboarding takes an optional workspace_name.

Fixed

  • “Continue with GitHub” only shows when GitHub sign-in is set up, so it never leads to a GitHub error page.

Upgrade

Self-hosted: pull the new image and restart. For GitHub sign-in set GITHUB_CLIENT_ID and GITHUB_CLIENT_SECRET: a GitHub App’s client ID and secret work, with “Email addresses: read-only” and <your url>/auth/github/callback as a redirect URI.

v1.0.0-beta.1

November 2, 2026

In short: ops is out, as a paid beta: deploy your apps to your own servers, and keep every change of their SQLite database in your own bucket. Hosted is $29 a month or $249 a year per workspace, 30 days free, no card.

New

  • Deploy: from your repo, a GitHub branch or commit, or an image from CI. Every push to GitHub can deploy by itself, and the commit shows the deploy status. A failed deploy brings the last good one back.
  • Backups: every SQLite change goes into your bucket, a snapshot before every change ops makes, restore to any second, and a restore check every day.
  • Data: browse tables, edit rows and run SQL on the live database.
  • Domains: HTTPS, a canonical domain, and a password in front if you want one.
  • Security: Cloudflare Tunnel instead of open ports, SSH only over Tailscale, a firewall, and a rules file on your server. The agent dials out; ops holds no key to your servers.
  • Teams and the activity log: roles, invitations, and who changed what, to the second.
  • Every surface: web, CLI, terminal UI, API and MCP.
  • Self-hosting: every feature, free, under AGPL.

Upgrade

New install: see getting started. What comes next is on the roadmap.