Not another chat rotator. A complete announcement engine.
Targeted, localized and event-aware broadcasts for modern Minecraft servers and networks.

_{6 runtimes  ·  8 notice surfaces  ·  8 audience rules  ·  3 event triggers  ·  3 languages shipped  ·  one render per language  ·  no database}

🧭 Why AutoMessage 📣 Broadcasts should earn attention ⚖️ More than a rotating list 🎯 Reach exactly who should hear it ⚡ Fire on an event, not a clock 🪶 Costs your server almost nothing 🌐 One engine, six runtimes

⚙️ Install and configure 📥 Get AutoMessage 🚀 Quick start 🛠 Build your first announcement 🗂 Configuration model 📋 Platform support 🔨 Building from source 💛 Support the project

📖 Reference 📡 Channels and selectors 🎯 Audience rules ⚡ Event triggers 🌍 Languages and fallback 🎨 MiniMessage and placeholders ⏱ Time format 💬 Discord webhook mirror ⌨️ Commands and permissions

📣 Broadcasts should earn attention

Players learn to ignore announcements when every message reaches everyone. AutoMessage makes server communication deliberate: every stream keeps its own schedule, every message chooses its own audience, and the text arrives in the language each player actually plays in - from one shared core running on Bukkit, Folia, Sponge, Velocity, Fabric and Minestom.

_{One announcement set, in game, on one server: a boss bar countdown, a full-screen title and a chat line.}

A complete announcement, start to finish

Greet returning veterans three seconds after they join - never on a timer, never the newcomers, and in their own language. Two files say all of it:

# scheduledMessages.yml - who gets it, and when
messages:
  - name: welcome-back
    trigger:
      type: JOIN
      delay: 3s
    rules:
      - type: PLAYTIME
        min: 10h

# lang/en.yml - what it says
announcements:
  welcome-back:
    - chat: "<gray>Welcome back, <aqua>{PLAYER}<gray>."
      sound: "entity.player.levelup MASTER 1.0 1.2"

And this is what the server does with them:

That is the whole feature set in miniature: an event instead of a clock, an audience instead of everyone, and text that lives beside every other translation.

⚖️ More than a rotating list

The tagline is not marketing. A rotator answers one question - what do I say next? - and sends the answer to everyone, in one language, on one clock.

	A plain rotator	AutoMessage
Who receives it	Everyone online	Permission, group, world, online count, playtime - composed with ANY_OF, NONE_OF, NOT
When it fires	One global interval	A schedule per channel, plus JOIN, FIRST_JOIN and re-arming PLAYER_COUNT triggers
What it looks like	A line of chat	Chat, action bar, title, subtitle, boss bar and sound, combined in one announcement
What it says	One text for everybody	The player's own language, with a per-message fallback
Where it runs	One server type	Bukkit, Folia, Sponge, Velocity, Fabric and Minestom, each a native artifact

The details underneath

🔀 Four rotation strategies SHUFFLE, SEQUENTIAL, RANDOM or weighted, chosen per channel. Shuffle deals a whole deck before anything repeats.	🧩 Configuration that fits the runtime A proxy never sees a world rule or a first-join trigger. Options the platform cannot honor are left out of the file, not written and ignored.	👁 Preview before publishing /automessage view <name> renders the real localized message to you alone, ignoring its audience rules.
↗ Beyond the server Mirror chat content to a Discord webhook, with player-only placeholders stripped and automatic mentions disabled.	♻ Reload, do not restart /automessage reload applies schedules, messages and languages in place - including a translation file added while the server was running.	🪶 Lean dispatch path No database. An empty server short-circuits, placeholders are scanned once per message rather than once per announcement, and everyone reading the same text in the same language is served from a single render.

🎯 Reach exactly who should hear it

Rules compose. Every top-level rule must pass, and ANY_OF, NONE_OF and NOT nest inside them when the audience needs more than an AND.

The same rules apply to a message sent by an event, so a join greeting restricted to a permission never reaches somebody who lacks it just because they walked through the door.

⚡ Fire on an event, not a clock

A message with a trigger leaves its channel rotation entirely and waits for something to happen.

PLAYER_COUNT is the one worth reading twice: it fires once on the way up and re-arms only after the server drops back below the threshold, so a population hovering on the boundary does not announce the same milestone every time somebody logs in.

🪶 Costs your server almost nothing

Formatting is the expensive part of an announcement. Colors, gradients, hover text and placeholders all have to be parsed before a single character reaches anybody - and the obvious way to do that is once for each player who receives it. Five hundred players online means the same sentence built five hundred times to say one thing, on the main thread, while the rest of the tick waits its turn.

AutoMessage builds it once per language and hands the finished result to everyone reading it.

	Formatting per player	AutoMessage
Parses per announcement	One per player, so 500	One per language actually being read
Time on the main thread	5.16 ms	0.34 ms
Share of a 50 ms tick	~10%, every announcement	~0.7%

That is 15× less main-thread time per announcement, measured on the real send path with 500 viewers online - not on a microbenchmark of one function. What remains is handing the finished text to each connection, which no plugin can avoid.

The saving grows with your player count, and it costs you nothing to configure: there is no cache to size, no setting to tune, and no behavior to give up.

It never trades correctness for speed. A message that reads differently for each player - one naming {PLAYER}, or carrying any PlaceholderAPI token - is still built for each of them, because only the expansion behind that token knows whether two players would see the same thing. Grouping applies exactly where the text is genuinely identical, and nowhere else.

Three more places the dispatch path refuses to do needless work:

• Placeholders are scanned once per message, not once per announcement. Which tokens a message contains is decided by the text, so it is worked out when the files are read and remembered until a reload - rather than re-derived every time the channel fires.
• An empty server short-circuits. Nobody online means no selection, no formatting and no work.
• Nothing touches disk while announcing. No database, no player records, no I/O on the hot path.

🌐 One engine. Six runtimes

AutoMessage uses a shared platform-neutral core and a small runtime adapter for each server type. That means the behavior stays familiar while scheduling, viewers, permissions and lifecycle hooks remain native to the platform underneath.

Artifact	Runtime	Supported line	Java	Platform-specific notes
Bukkit	Spigot, Paper, Purpur and other Bukkit-derived servers	Minecraft 1.21–26.2	21+	Full feature set; optional PlaceholderAPI integration
Folia	Folia	Minecraft 1.21–26.2	21+	Dedicated regionized scheduling; optional PlaceholderAPI integration
Sponge	Sponge API 17	Minecraft 1.21.10	21+	Full feature set except PlaceholderAPI
Velocity	Velocity 3.5.1 proxy	Proxy runtime	21+	No world, playtime, first-join or PlaceholderAPI capabilities
Fabric	Dedicated Fabric server	Minecraft 1.21.11	21+	Server-side mod; Fabric API required; no PlaceholderAPI
Minestom	Embedded Minestom library	Protocol 1.21.11	25+	Classpath library; no world, playtime, first-join or PlaceholderAPI; permission/group rules require your callback

Note

Java 21 is the bytecode floor for the plugin artifacts, not a promise that every Minecraft server version runs on Java 21. Use the JVM required by your server; Minestom itself requires Java 25.

Each artifact writes a platform-aware configuration. Options the runtime cannot honor are left out instead of being generated and silently ignored.

What about Quilt?

There is no separate Quilt artifact. Quilt can read Fabric metadata, but this branch does not claim a tested Quilt version. Treat the Fabric jar on Quilt as experimental and bring a compatible Fabric API yourself.

🗂 Configuration model

AutoMessage separates when, who and what. Adding a language never requires copying scheduling rules, and changing a channel interval never touches the message text.

flowchart LR
    schedule["config.yml<br/>WHEN<br/>channels · periods · selectors"]
    audience["scheduledMessages.yml<br/>WHO<br/>rules · weights · triggers"]
    content["lang/*.yml<br/>WHAT<br/>localized MiniMessage notices"]
    minecraft["Minecraft<br/>PLAYER OUTPUT"]
    discord["discordWebhook.yml<br/>OPTIONAL MIRROR"]

    schedule --> audience --> content --> minecraft
    content -. "chat text" .-> discord

    classDef config fill:#11182d,stroke:#8b7cf6,color:#eef0ff,stroke-width:1px;
    classDef output fill:#0d2028,stroke:#45d7e8,color:#ecfeff,stroke-width:1px;
    class schedule,audience,content config;
    class minecraft,discord output;

Loading

File	Responsibility
config.yml	Master switch, fallback language, channels, startup delays, periods and selectors
scheduledMessages.yml	Message names, channel assignment, weights, audience rules and event triggers
lang/<code>.yml	Localized command replies and the actual MiniMessage announcement payloads
discordWebhook.yml	Opt-in Discord webhook, display name and avatar

📥 Get AutoMessage

What you need	Where to get it
Current public Bukkit build	Modrinth · SpigotMC · Hangar
Release history and changelogs	GitHub Releases
Multiplatform artifacts from this branch	Build all six from source

Important

Release status: this README documents the local multiplatform branch. The current public listings still represent the earlier Bukkit-focused release line. Until a new GitHub release publishes all six artifacts, build the multiplatform jars with ./gradlew dist.

🚀 Quick start

1. Get the artifact built for your runtime. Platform jars are intentionally separate.
2. Put it in your platform's plugins/ or mods/ directory. For Minestom, add the jar to the application classpath and use the builder shown below.
3. Start the server once. AutoMessage generates annotated configuration plus working example announcements.
4. Edit config.yml, scheduledMessages.yml and the files inside lang/.
5. Run /automessage reload, then test a message with /automessage view <name>.

Tip

Start with the generated files. They are not empty templates: every available field is explained next to a working example, and the contents adapt to the platform that created them.

🛠 Build your first announcement

The example below creates an independent event stream, adds one announcement to it, and renders multiple notice types from a single localized payload.

1. Schedule the stream

# config.yml
enabled: true
fallbackLanguage: en

channels:
  - name: default
    enabled: true
    initialDelay: 1m
    period: 5m
    selector: SHUFFLE

  - name: events
    enabled: true
    initialDelay: 30s
    period: 15m
    selector: SEQUENTIAL

2. Register the announcement

# scheduledMessages.yml
messages:
  - name: event-night
    channel: events

3. Design what players receive

# lang/en.yml
announcements:
  event-night:
    - chat:
        - "<dark_gray>[<gold>EVENT<dark_gray>] <gray>The arena opens in <gold>5 minutes<gray>."
        - "<gray>Click <click:run_command:'/warp arena'><hover:show_text:'<yellow>Teleport'><aqua><u>here</u></aqua></hover></click> to join."
      actionbar: "<yellow>Prepare your loadout - the event starts soon"
      title: "<gradient:#ffd166:#ff6b6b><bold>EVENT NIGHT</bold></gradient>"
      subtitle: "<gray>The arena opens in 05:00"
      times: "500ms 3s 500ms"
      bossbar:
        message: "<light_purple>Event countdown"
        duration: 10s
        color: PURPLE
        overlay: PROGRESS
        progress: 0.75
      sound: "block.note_block.pling MASTER 1.0 1.0"

Use only the parts that fit the message. A simple chat announcement can be a single string; a showcase announcement can combine every surface above.

📖 Reference

Announcement channels and selectors

Every channel has its own schedule and independent selector state:

channels:
  - name: tips
    enabled: true
    initialDelay: 1m
    period: 5m
    selector: SHUFFLE

  - name: promotions
    enabled: true
    initialDelay: 2m
    period: 20m
    selector: WEIGHTED

Selector	Behavior
SHUFFLE	Randomizes the order and shows every message once before anything repeats
SEQUENTIAL	Walks straight through the configured list
RANDOM	Draws independently; the same message may appear twice in a row
WEIGHTED	Draws randomly in proportion to each message's weight

SEQUENTIAL and SHUFFLE keep their position across a normal configuration reload. A weight of 0 parks a message only in a WEIGHTED channel; other selectors do not consult weights.

Audience rules

Top-level rules behave like logical AND. Nest ANY_OF, NONE_OF and NOT when the audience needs more expressive logic.

Rule	Matches
PERMISSION	Viewers holding a permission node
GROUP	Viewers holding group.<name>; this is permission-based, not a direct LuckPerms/Vault lookup
WORLD	Viewers in one of the configured worlds
PLAYER_COUNT	While the live online count is inside a configured range
PLAYTIME	Viewers whose native platform playtime is inside a configured duration range
ANY_OF	At least one nested rule
NONE_OF	None of the nested rules
NOT	The inverse of one nested rule

messages:
  - name: vip-event
    channel: events
    rules:
      - type: ANY_OF
        rules:
          - type: PERMISSION
            permission: rank.vip
          - type: PERMISSION
            permission: rank.staff
      - type: PLAYER_COUNT
        min: 10
        max: 100

Rules still apply to messages sent by an event trigger. Platform-specific rules are generated only where the runtime can supply the required data.

Event triggers

A triggered message leaves its timed channel rotation completely:

messages:
  - name: welcome-back
    trigger:
      type: JOIN
      delay: 3s

  - name: first-join-welcome
    trigger:
      type: FIRST_JOIN
      delay: 3s

  - name: one-hundred-online
    trigger:
      type: PLAYER_COUNT
      threshold: 100

PLAYER_COUNT fires once when the threshold is reached and re-arms only after the online count drops below it. FIRST_JOIN is unavailable on runtimes that do not expose a persistent first-join signal.

[!NOTE] /automessage enable and /automessage disable control scheduled channels. Event-triggered announcements operate independently.

Languages and fallback behavior

AutoMessage ships en, pl and de. Add another language by copying a file to lang/<code>.yml, translating it and running /automessage reload.

For a player using pt_br, AutoMessage looks for:

1. lang/pt_br.yml
2. lang/pt.yml
3. the fallbackLanguage from config.yml

A missing announcement falls back individually, so a new translation does not have to be complete before it is useful.

MiniMessage, placeholders and notification parts

All text uses Kyori MiniMessage, including colors, gradients, decorations, hover text and click actions.

Built-in placeholders:

Player-scoped	Server-scoped
{PLAYER} · {DISPLAY_NAME} · {UUID} · {WORLD}	{ONLINE} · {MAX_PLAYERS} · {DATE} · {TIME}

On Bukkit and Folia, an installed PlaceholderAPI is detected automatically. Standard %placeholder_name% tokens containing letters, digits and underscores can then be used alongside the built-ins.

Available notification keys:

Key	Value
bare string / chat	One chat line or a list of lines
actionbar	MiniMessage text shown above the hotbar
title / subtitle	Title layers; either can be used independently
times	Three durations: fade in, stay, fade out
hideTitle	Explicitly clear the current title
bossbar	Message, duration, color, overlay and optional progress 0.0–1.0
sound	Sound key, or key source volume pitch

Need a visual starting point? Try the external EternalCode Notification Generator and adapt the generated notification to the language-file structure above.

Time format

Every duration accepts an explicit unit, and units can be combined:

Unit	Meaning	Example
ms	milliseconds	500ms
s	seconds	30s
m	minutes	5m
h	hours	2h
d	days	1d

1m30s is valid. A plain number is interpreted as seconds. Channel periods shorter than 50ms are normalized, and tick-based runtimes ultimately execute on tick boundaries.

Discord webhook mirror

Set enabled: true and provide a Discord webhook URL in discordWebhook.yml to mirror scheduled and triggered announcements.

The integration is deliberately conservative:

• only chat parts are mirrored;
• MiniMessage is flattened to plain text;
• the configured fallback language is used;
• server placeholders are resolved, while player-scoped values are removed;
• automatic Discord mentions are disabled;
• the webhook must use HTTPS and a Discord webhook host/path;
• rate-limit responses are retried with backoff.

The webhook is opt-in and makes no outbound request while disabled or missing a valid URL. Restart AutoMessage after changing the webhook's enabled state or URL so the connection is rebuilt.

Commands and permissions

Command	Permission	Description
/automessage enable	command.automessage.enable	Resume scheduled announcement channels
/automessage disable	command.automessage.disable	Pause scheduled announcement channels
/automessage reload	command.automessage.reload	Reload schedules, message definitions and languages
/automessage view <name>	command.automessage.view	Preview one localized message; player-only, with tab completion
/automessage next	command.automessage.next	Show what each channel will announce next and how long until it fires, without using it up
/automessage send <channel>	command.automessage.send	Send that channel's next announcement now; its interval restarts from this moment
/automessage toggle	command.automessage.toggle	A player turns announcements off for themselves, or back on; remembered across restarts
/automessage stats	command.automessage.stats	Per channel: how much it has carried, when the last one went out and when the next is due, then the same per message

Embedding AutoMessage in Minestom

Minestom has no plugin directory. Initialize it, place AutoMessage on the classpath and enable the library after Minestom itself is initialized:

MinecraftServer.init();

AutoMessageMinestom autoMessage = AutoMessageMinestom.builder()
        .dataDirectory(Path.of("automessage"))
        .permissions((sender, node) -> myPermissions.check(sender, node)) // optional
        .enable();

// During your server shutdown:
autoMessage.shutdown();

Without a permission callback, AutoMessage uses Minestom's operator level to protect its commands. Permission and group audience rules are then omitted because the runtime cannot answer them accurately.

🔨 Building from source

Building every runtime requires JDK 25. From the repository root, run:

./gradlew dist

The task runs all checks, builds each platform-specific jar and synchronizes the six installable artifacts into build/dist:

• Bukkit
• Folia
• Sponge
• Velocity
• Fabric
• Minestom

If any verification fails, the distribution is not produced. A normal ./gradlew build performs the same project checks but leaves artifacts inside their individual modules.

💛 Support the project

AutoMessage is free, open source and maintained independently.

Found a problem?	Want to contribute?	Want to support development?
Open a bug report or feature request	Read the contribution guide and open a pull request	GitHub Sponsors · PayPal

Please follow the project's Code of Conduct in all community spaces. AutoMessage is distributed under the GNU General Public License v3.0.

If AutoMessage makes your server communication better, leave the repository a star.
It helps other server owners discover the project.