Seylo
Star on GitHub
NewDeploy to many nodes from one panel

All your servers.
One calm panel.

Deploy apps, databases and compose stacks to as many Linux servers as you own. Push to GitHub, Seylo builds it, checks it, and switches traffic without downtime.

Starts the panel and makes this server its first node. Add more servers from the panel.

Get startedSee what it does
  • Open source
  • AGPL-3.0
  • Self-hosted
The panel's dashboard: node load graphs for three servers and the projects running on themThe panel's dashboard: node load graphs for three servers and the projects running on them
Features

Everything a small platform team would build. Already built.

One panel, any number of nodes, and the boring parts handled: certificates, restarts, cleanup, updates.

  • deploy

    Push to deploy, zero downtime

    GitHub, any Git host, Docker images or a Dockerfile. A failed health check keeps the old version live.

  • nodes

    Many servers, one panel

    Germany, the US, your closet. Each node keeps running its apps even when the panel is down.

  • data

    Databases that stay private

    Postgres, MySQL, MariaDB, Mongo and Redis on a private network, reachable by name. Expose one with SSL when you need to.

  • compose

    Compose stacks and templates

    Paste a compose file or install from the catalog. Every container gets logs, domains and restarts.

  • domains

    Domains on autopilot

    A wildcard per node gives every service a domain. Custom domains get HTTPS on their own.

  • alerts

    Crash reports that say why

    Restarts on crash, the last log lines in your inbox, Discord or any webhook.

  • console

    Console, shell and files

    Type into the running app, open a shell, browse volumes, all from the browser.

  • mcp

    Built for AI assistants

    An MCP server lets Claude deploy, read logs, change env and open a dev tunnel. Read-only by default; deletes wait for your approval.

  • brand

    Yours to brand

    White label, theme presets, custom CSS and your own maintenance pages.

Templates

23 apps, one click each.

n8n, Plausible, Grafana, Vaultwarden and the rest install as normal services: a pinned image, a database when it needs one, a domain and its own passwords. Change anything afterwards.

  • See the services before anything is created
  • Passwords generated for you
  • Template code runs sandboxed
How templates work
The Templates page: a search for 23 templates, category filters and cards for Appsmith, code-server, Directus, Excalidraw, File Browser and GhostThe Templates page: a search for 23 templates, category filters and cards for Appsmith, code-server, Directus, Excalidraw, File Browser and Ghost
  • Appsmith
  • code-server
  • Directus
  • Excalidraw
  • File Browser
  • Ghost
  • Gitea
  • Grafana
  • IT Tools
  • listmonk
  • Mattermost
  • Meilisearch
  • Metabase
  • n8n
  • NocoDB
  • pgAdmin
  • Plausible Analytics
  • Prometheus
  • Redis Insight
  • Stirling PDF
  • Umami
  • Uptime Kuma
  • Vaultwarden
  • Your own
Per service

Settings you'd otherwise script.

What usually lives in a Traefik file, a cron job or a wiki page is a field on the service, applied by the node.

  • Env that knows your other services

    ${{redis.INTERNAL_URL}} resolves on every deploy. The editor underlines a reference this node can't reach and offers the fix. Hover one to see its value.

    The environment editor: two lines underlined in red with one-click fixes below them such as Expose redis, and the hover on ${{api.URL}}: api's public URL becomes https://api.de1.example.com and updates if the domain changesThe environment editor: two lines underlined in red with one-click fixes below them such as Expose redis, and the hover on ${{api.URL}}: api's public URL becomes https://api.de1.example.com and updates if the domain changes

  • Access rules per domain

    Basic auth, an IP allowlist, a rate limit and regex redirects on each domain. No middleware files to edit.

    Access rules for one domain: basic auth users, an IP allowlist, a rate limit of 20 requests per second and a redirectAccess rules for one domain: basic auth users, an IP allowlist, a rate limit of 20 requests per second and a redirect

  • Health checks that gate deploys

    TCP, an HTTP request with path, host and expected status, or a command. A new version gets traffic only after it passes; if it never does, the old one keeps serving.

    Health check settings: an HTTP request to /health on port 3000, interval, timeout, grace period and the zero-downtime deploy strategyHealth check settings: an HTTP request to /health on port 3000, interval, timeout, grace period and the zero-downtime deploy strategy

  • A crash policy, not a restart loop

    Restart after a fixed 5 seconds, stop after N crashes in a row, and get one alert per streak instead of one per crash.

    Crash policy: restart after a crash is on, stop after 5 crashes in a rowCrash policy: restart after a crash is on, stop after 5 crashes in a row

  • Timed expose, locked to your IP

    A port open for 5 minutes to 24 hours, only to the IPs you list. The node closes it on time, even with the panel down.

    An open timed expose: a node port to web:3000, only from one IP, closing in 1 hour 41 minutesAn open timed expose: a node port to web:3000, only from one IP, closing in 1 hour 41 minutes

  • Maintenance and down pages

    One switch puts one service, or all of them, behind a maintenance page that answers 503. A down page shows by itself while a service is stopped, on its first deploy or unreachable.

    • Built-in page with your message
    • Your own HTML, scripts off
    • Per service or panel-wide
Nodes

Many servers, one panel.

Add a server with one command from the panel. Each node keeps its services running, restarting and routing on its own, even while the panel is down.

Live graphs for every node

CPU, memory, disk and network, live or over 30 days, next to the machine's facts and the services it runs. Nodes dial out to the panel, so they need no open port for it.

  • Live to 30 days
  • Machine facts
  • Agent logs
  • Reboot
A node's overview: CPU, memory, disk and network graphs, and the machine's OS, kernel, Docker and agent versionsA node's overview: CPU, memory, disk and network graphs, and the machine's OS, kernel, Docker and agent versions

Replicas across nodes, or a clean move

Run copies on several nodes. Each node's edge spreads requests over all replicas, or pick an entry node that takes the public traffic and does the spreading. Moving a service deploys it on the new node, stops the old copy while its volumes are copied over, switches traffic, then removes the old copy. If anything fails, it stays where it was.

  • Load balanced
  • Entry node
  • Sticky sessions
  • Volumes move with it
Replicas and nodes: two replicas on de-fsn-1 and one on us-ash-1, de-fsn-1 as the entry node, sticky sessions onReplicas and nodes: two replicas on de-fsn-1 and one on us-ash-1, de-fsn-1 as the entry node, sticky sessions on

A health check with the fix

16 checks per node: kernel, disk, Docker, cgroups, ports, proxy, DNS, firewall, clock, the panel connection and more. Each problem comes with the step that fixes it. Updates roll the panel first, then one node at a time, and a node that fails to start the new version rolls back by itself.

  • 16 checks
  • Auto-update
  • Panel first, then node by node
  • Rollback
A node's health check: one warning about disk space with its fix, and passed checks for the kernel, OS, memory, inodes, Docker and cgroupsA node's health check: one warning about disk space with its fix, and passed checks for the kernel, OS, memory, inodes, Docker and cgroups
Databases

Five engines, private by default.

Each project has its own network. Services on the same node reach a database by name; nothing else can until you expose it.

  • PostgreSQL
  • MySQL
  • MariaDB
  • MongoDB
  • Redis
A Postgres service: the internal connection, an exposed connection that requires SSL, pgweb and psql under Tools, and the daily backupsA Postgres service: the internal connection, an exposed connection that requires SSL, pgweb and psql under Tools, and the daily backups

  • Exposed over TLS only

    Exposing one adds a separate user that must use TLS; Redis keeps its default user behind TLS. The internal URL stays the same.

  • A web UI, signed in

    pgweb, phpMyAdmin, mongo-express or Redis Commander, started on click and stopped when idle.

  • A shell that's logged in

    psql, mysql, mariadb, mongosh or redis-cli in the browser, no password to look up.

  • Backups to S3 or R2

    Dumps on a schedule with keep-last retention, to S3, R2 or MinIO. Restore any kept run.

Logs, metrics, alerts

See what your apps are doing.

Every service gets logs, metrics, uptime checks and crash reports. No second tool to install.

A service's logs: a search across every deploy's logs above the live log, following new linesA service's logs: a search across every deploy's logs above the live log, following new lines

Logs across deploys

Search every deploy's logs, not just the running one, or follow them live. The console, a shell and a file browser for the volumes open in the browser too.
A service's CPU over the last hour, average and peakA service's CPU over the last hour, average and peak

Metrics per service

CPU, memory and network, average and peak, from live up to 30 days.
An uptime check that is up, 99.93% in 24 hours, with its latency chart, and recent alerts such as worker crashed: exit 137, out of memoryAn uptime check that is up, 99.93% in 24 hours, with its latency chart, and recent alerts such as worker crashed: exit 137, out of memory

Uptime checks and crash reports

A URL checked on an interval, with its latency history. A crash alert carries the exit code and the last 100 log lines.

Alerts where you already look

Choose which events go where: crashes, failed builds and deploys, health checks, uptime, certificates, backups, cron runs, disk space, offline nodes. Panel-wide defaults, changeable per service.

  • Email
  • Discord
  • Slack
  • Telegram
  • ntfy
  • Pushover
  • Gotify
  • Webhooks
People and access

Let people and tools in, one scope at a time.

  • Roles and grants

    Owner, admin and member. Members reach only the projects and services you grant them, down to single actions.

  • 2FA and passkeys

    Authenticator codes with recovery codes, and passkeys. New keys and sign-in changes ask you to confirm it's you.

  • Scoped API keys

    Limited to the actions and projects you pick, read-only if you like, with an expiry date.

  • An audit log

    Every change through the panel, the API, MCP or the assistant, and every refused attempt. Filter it, export it.

  • A CLI for your laptop

    Login opens the panel in your browser: type the code your terminal shows, pick what the key may do, approve.

    seylo login https://panel.example.com

  • MCP with an approvals queue

    AI tools start read-only. Deletes and anything that can't be undone wait in Approvals until a person says yes; the rest runs only with the actions you gave the token.

An approval waiting: an MCP client asks to delete a project, with its reason and exact input, and Deny or Approve and runAn approval waiting: an MCP client asks to delete a project, with its reason and exact input, and Deny or Approve and run
People: an owner with 2FA and two passkeys, an admin, a member with access to two projects and services, and a disabled memberPeople: an owner with 2FA and two passkeys, an admin, a member with access to two projects and services, and a disabled member
White label

Make it yours.

Hand the panel to a team or to clients under your own name. Four theme presets, the default one following light or dark mode, then your accent color and your own CSS on top.

  • Panel name
  • Logos for light and dark
  • Favicon
  • Sign-in message
  • Hide "Powered by"
  • Accent color
  • Custom CSS
Appearance settings: four theme presets, accent colors, branding with the panel name, logos, favicon and sign-in message, and custom CSSAppearance settings: four theme presets, accent colors, branding with the panel name, logos, favicon and sign-in message, and custom CSS
  • The dashboard in the Default theme
    DefaultLight and dark
  • The dashboard in the Night theme
    NightDark with ink blues
  • The dashboard in the Classic theme
    ClassicSoft gray cards, green accent
  • The dashboard in the Midnight theme
    MidnightDense, purple accent
Already running Pterodactyl or Pelican?

Runs next to your game servers. No extra VM.

Seylo shares the node's proxy with Wings and never touches containers it didn't create. One switch gives game servers first claim on the CPU, so a big build can't cause lag.

  • il-tlv-1Wings detected
  • cs2-retake-01, cs2-mm-04priority
  • api, worker, umamishare the rest
Get started

Own your platform.

Free and open source. Bring a server with Docker.

Read the install guideView the source