📋 Server Assistant Changelog — v6.x

/meme now reads like an emote. Instead of just echoing your search, the bot adds a caption describing what the meme says about you — /meme this is fine → “@User has decided everything is fine”, /meme drake → “@User has notes”, /meme woman yelling at cat → “@User is losing an argument to a cat”.

• Smart, in-context captions — on AI-enabled servers (your AI allowance, free trial included — not Premium-only) the caption is written to fit the meme and the conversation, accurate and a little funny. It’s governed by the same AI theme-pick toggle as emotes in /settings → Emotes.
• Always works — when AI isn’t available it falls back to a built-in library of well-known memes for a witty caption, and an unrecognised meme just posts the image. No dead ends.

GIFs & memes by KLIPY.

Time spent together in voice chat now counts toward your level. While you’re actively in a voice channel, you earn XP for it — feeding the same balance as chatting. The busier the channel, the more it’s worth: a lively call with several people pays more per minute than a quiet two-person chat.

• It rewards being present, not just connected. You earn while you’re actually in the conversation — sitting muted to listen on a movie or study night still counts. You won’t earn while you’re parked in the server’s AFK channel, deafened, or alone in a channel, so nobody can farm levels by idling overnight.
• Double-XP events apply here too. If your server’s running a Double-XP event (or you’ve got a personal Double-XP buff from leaving a review), your voice time earns at 2× as well.
• Free for everyone. Like chat XP, voice XP is on for all servers — no Premium needed.

Drop the perfect meme into chat. New /meme <search> posts a meme image matching your search — /meme distracted boyfriend, /meme this is fine, /meme success kid. Memes come from KLIPY.

• Shares your emote controls. The same /settings → Emotes panel governs memes — channel allowlist, per-minute rate limit, and NSFW safe-search all apply — plus a new Memes on/off toggle (on by default).

GIFs & memes by KLIPY.

Emotes now show what you’re actually doing. /laugh posts a laughing GIF, /cry a crying one — the search leads with the action and prefers clips whose title or tags match it, so you get a relevant reaction instead of a generic theme image. Add a theme on top (/cry anime) and it still applies; if nothing in that theme fits the action, you get a plain matching GIF and a small note saying so.

• Captions read like a sentence. Posts now say @User is crying, @User hugs @Target, or @User waves at @Target — instead of just the emote’s name.

GIFs by KLIPY.

React with a GIF in one tap. New /emote (and the shortcut /e), plus /cry, /hug and /dance, drop a fitting GIF straight into chat — with autocomplete over hundreds of emotes (cry, dance, laugh, facepalm, wave… and aliases like lol, ty, congrats).

• Smart by default. Just run /cry and the bot picks a theme that fits your server and the moment, then grabs a random matching GIF — so it’s fresh each time. Prefer your own flavour? Add it: /cry anime, /cry wow, /cry naruto.
• Make it yours. /emotestyle anime sets your personal default theme so your emotes lean your way across every server; /hug @member aims a reaction at someone.
• Powered by your AI allowance — free trial included. The auto-theme uses your server’s AI (the free 150K trial counts); once that’s used up it simply falls back to a plain random GIF, so the commands always work. It’s not Premium-only.
• Server controls in /settings → Emotes. Turn emotes on/off, limit them to certain channels, set a per-minute rate limit, allow spicier GIFs only in NSFW channels, and toggle an off-by-default “flavor pack.” Safe-search is on by default.

GIFs by KLIPY.

The “you now have **X XP” line in vote and review thank-you messages now matches Crestbound.** Your wallet has two numbers: the lifetime XP that ranks you on the leaderboard (it only ever goes up), and your spendable XP — the balance you spend on Booster Packs in Crestbound, which drops each time you open a pack. Both go up together when you earn, so they match until you spend.

Those reward messages were showing the lifetime number, so after you’d opened a pack it could read higher than the balance Crestbound actually shows. Since the message is telling you to go spend it, it now quotes your spendable balance — the same figure you see in Crestbound. Nothing changed about how much you earn or can spend; only the number in the message.

Leave us a review, get rewarded. Run /review to get your personal code, add it to a review you write on a bot directory, then run /review with the link — and you’ll earn 500 XP plus 1 month of Double XP.

• Double XP that follows you. For a month, you earn XP twice as fast in every server you share with Server Assistant — then spend it in Crestbound. Earn it again on another directory and the month is added on, not wasted.
• Across the directories — Top.gg, DiscordForge, Discord Bot List and discord.bots.gg. One reward per platform.
• Verified by your code, not a screenshot. The bot checks that your unique code is present on the live review page — so it can’t be faked, and only your own review counts. If a site hides reviews from automated checks, your link goes to our team for a quick manual confirmation and you’re rewarded by DM.

Thanks for helping more communities find the bot! 💚

ThreatNet is now live. When a user with a serious, corroborated history of abuse on other protected servers joins yours, your staff now see a clear advisory flag on them — free, on every server — and decide what to do. The flag never bans or sanctions anyone on its own; your team stays in control.

• Premium servers can go further with ThreatNet auto-protect. Switch it on and the bot will automatically ban a joining user whose account meets a cross-server risk threshold you choose — closing the gate before they can act. It’s off by default; turn it on with /threatnet autoban on, or in /settings → Security → ThreatNet. The safest high threshold (serious and corroborated across two or more independent servers) is the default; you can set a broader one, and you choose the level.
• Privacy-first and severity-only. Only a minimized, generic severity signal is ever shared across servers — never the offence type, your staff’s reasons, AI summaries, or message content. Individuals can opt out of profiling at any time. Full detail is in our Privacy Policy.
• Never acts on opt-outs. Auto-protect never bans anyone who has opted out of network profiling.

Setting up verification just got a lot easier. Verification works by hiding your channels from people who haven’t been verified yet and revealing them once they pass — but wiring that up by hand, channel by channel, is tedious.

Now there’s a “Auto-configure channels” button in /settings → Members → Verification. It proposes which channels stay public (your verify channel, plus rules / announcements), lets you keep any extras visible with a quick picker, and then — in one tap — hides everything else from unverified members and opens it to your verified role.

• Nothing is guessed blindly — you see exactly what will change before you confirm.
• One-tap Undo — it snapshots your current setup first, so you can revert instantly if it’s not what you wanted.
• You’ll need a verified role set and the bot’s Manage Channels/Roles permission.

Also: the Notifications settings menu is flatter — it opens straight to your notification options (with Instant alerts and Pulse inside), instead of an extra step.

/settings is tidier. As Server Assistant has grown, the settings list got long — so it’s now grouped into four sections:

• ⚙️ Settings — branding, presets, behavior, role tiers, AI, white-label, snippets, custom commands, FAQ
• 👥 Members — new-member handling, verification, onboarding, role panels
• 🔔 Notifications — event pings and the daily Pulse digest
• 🛡️ Security — AutoMod, anti-nuke, scam-image protection, channel allowlist, privacy, backups

Tools that used to only have their own command — like AutoMod, AI config, onboarding, backups and more — can now also be opened straight from the matching section of /settings. Their commands still work exactly as before, and nothing about who can use what has changed.

Sick of seeing scams like this? An account joins, posts a screenshot of a fake “withdrawal received” or a too-good-to-be-true trading profit, and vanishes — and because the whole pitch is inside a picture, ordinary word filters never see it.

ThreatNet now recognises scam images. The moment a scam screenshot is flagged anywhere on the network, every server is protected from it — Server Assistant spots the same image and removes it automatically, before your members fall for it.

• It just works — on by default, for everyone, on every plan. Nothing to set up. Known scam images are removed the moment they’re posted, and the person who posted one is timed out.
• It catches brand-new scams too. On AI-enabled servers, images dropped by brand-new members are checked for never-seen-before scams — and once spotted, the whole network learns to block them.
• You’re in control. Prefer to review instead of auto-remove? Switch it to quarantine or flag-only under /settings.
• Privacy-first. Servers only ever share a small, irreversible fingerprint of a flagged scam image — never the image itself, and never anything tied to a person. The same opt-out that covers the rest of ThreatNet covers this too.

It’s the biggest upgrade to ThreatNet yet — turning every server that flags a scam into protection for all the others.

There’s now a full wiki for Server Assistant — a complete, browsable reference that shows a live mock-up of exactly what you’ll see in Discord for every command, wizard and flow.

• Main features — a guided tour of what the bot does, with deep-dive pages for Pulse, ban appeals, the audit log, alt / ban-evasion detection and ThreatNet.
• Every command — grouped by area (moderation, AutoMod, AI, tickets, member experience, account & premium), each shown the way it appears in Discord.
• Step-through wizards — walk the /setup wizard and other multi-step flows one screen at a time.
• Search — find any command or feature in seconds.

Open it from Main Features in the top menu, or jump straight in at the wiki. Setup and the full command reference now live there too.

You can now bring your own Anthropic (Claude) key. Anthropic has long been listed as a supported bring-your-own-key provider, but the key-entry form only accepted OpenAI and xAI keys — so an sk-ant-… key was turned away. That’s fixed: in /setup and /ai-config, the form now accepts your Anthropic key alongside OpenAI and xAI.

Just paste the key — the provider is auto-detected from it (sk-ant-… → Anthropic, sk-… → OpenAI, xai-… → xAI), so the provider box stays optional, and you can leave the model blank for a sensible default. As with any BYOK key, your server runs AI on your own account (no trial limit) once you’re on Premium BYOK ($3/month).

A smoother “use my own key” step in /setup. If you chose 🔑 Enter my own key and then closed the form, or typed the provider in a way the bot didn’t recognise, the wizard could get stuck — every Finish said “you didn’t fill the form” with no obvious way forward, even when you’d pasted a perfectly good key.

Now the bot auto-detects the provider from your key (an sk-… key is OpenAI, an xai-… key is xAI), so the provider box is optional. If you change your mind or close the form, the included free trial (150k tokens) stays selected and Finish just works. And if you paste a key the bot can’t use here, it tells you plainly — the built-in AI already runs on Claude, so no key is needed for the free trial.

A smoother permission review during /setup. When the wizard asks you to review permission changes, applying them can take a few seconds — a series of role and channel edits. Previously, on a slow apply, the Approve & Apply, Re-check and Generate fresh buttons could appear to do nothing, leaving you to click again.

Those buttons now acknowledge your click straight away and then apply the changes, so the wizard moves on reliably no matter how long the apply takes.

A clearer first run. If you add Server Assistant and start using commands before running setup, they used to quietly do nothing — which can look like the bot is broken.

Now, in a server that hasn’t been set up yet, commands point you to /setup so it’s obvious what to do first. The getting-started and personal commands still work right away — /setup, /help, /invite, /portal, /whatsnew, /support, /premium, plus your account ones like /rank, /vote and /leaderboard (your XP follows you, not a single server, so those work everywhere). Everything else unlocks the moment setup is complete.

Your members earn XP just by being part of the conversation — and now there’s a whole game to spend it on.

Server Assistant quietly tracks how active your members are: every message earns a little XP, members climb through levels, and a server leaderboard shows who’s most involved. It’s built in, on by default, and needs nothing to set up — see XP & leveling for /rank, /leaderboard and level-up announcements.

🎴 Crestbound — the headline. The XP you earn is a currency you can spend in Crestbound, the Crest-collecting game in the web portal:

• Open booster packs with your XP and collect the Crests of the first Chapter, Embers of the First War, across four rarities.
• Discover brand-new Crests. The first person to open an undiscovered Crest summons its artwork — drawn once and then shared with every player, forever, with a line of credit to the discoverer.
• The grand reveal. When your batch finishes illuminating, one ✦ Unveil the Crests moment turns them all over at once — the heralds present your Crests, art and all, seen for the very first time.
• Burn spare Crests back into XP, and complete your collection.

⚡ Double-XP events. When a server has gone quiet, the bot offers your staff a one-tap 48-hour Double-XP event — while it runs, every member earns 2× XP for chatting. Only a server manager can start one, and the staff chat gets a recap of the total XP earned when it ends.

🗳️ Better vote rewards. A vote for Server Assistant on DiscordForge now earns 75 XP — doubled to 150 if you’re in a server that has SA — and you can vote again every 8 hours. The thank-you and reminder DMs got a refresh, and the Crestbound page shows a live countdown to your next vote.

📊 Weekly recap. Once a week, if you earned any XP, the bot sends you a friendly DM with your total — split into how much came from being active versus voting.

A note on fairness: XP is tied to you, not a single server — chat XP from every server you’re in, plus your voting rewards, add up into one account-wide total that you spend in Crestbound.

The second feature you voted for in v6.0 is here — verification that only gets in the way of suspicious joins.

Server Assistant now scores every new member as they join and decides whether they need a quick check. There are three modes:

• Intelligent (the new default) — regular members walk straight in. Only risk-scored joins — brand-new accounts, no profile picture, or accounts carrying cross-server or ban-evasion signals — are held for a quick check. The right balance of safety and zero friction for real members.
• On — everyone completes verification before they get access.
• Off — no verification.

When someone is held, what happens next depends on how risky they look:

• Lower-risk → self-serve. They get a link to a quick web page: sign in with Discord, confirm you’re a real person, done — access is granted automatically, no staff needed. They’re pointed there from a tidy, read-only #verify channel.
• Higher-risk → your call. Your staff get a one-tap Approve / Kick / Ban card with the reasons, so a human decides. You set where that line sits.

A few things worth knowing:

• It’s tuned to be safe by default. New servers start on Intelligent. Existing servers are only switched on automatically if you already had a verified-access role set up — otherwise it stays off until you choose to turn it on. You’ll get a one-time heads-up in your staff channel either way.
• Everything stays on your server — verification answers and decisions are never shared anywhere.
• Manage it all under /settings → Verification — the mode, how sensitive it is, and the self-serve cut-off.

Also in this release: an unban now clears the slate — if you unban someone (or lift a ban directly in Discord), it no longer counts against them in alt-detection or join risk-scoring. An overturned ban shouldn’t follow someone around.

One of the three features you voted for in v6.0 is here — protection against a server “nuke” or a compromised admin account.

Server Assistant now watches for the tell-tale signs of a server nuke: one account suddenly deleting channels or roles, mass-banning or kicking members, or grabbing dangerous permissions in a tight burst. When that happens, it works out how likely it is to be an attack and responds straight away:

• It stops the damage. At high confidence it automatically quarantines the offending account — stripping the roles giving it power — then alerts you and your staff with a short, plain-language summary of what happened. If it was a false alarm, one tap puts everything back.
• You’re always safe. The server owner is never affected, and you can add trusted staff, roles or bots to a whitelist so their normal admin work is never flagged.
• Tuned to your community. Sensitivity is set automatically from your server type (busier servers get more headroom; locked-down servers are stricter), and you can adjust it yourself.
• Put your server back exactly. Server Assistant keeps a regular snapshot of your roles’ permissions and channel settings, so it can restore them precisely — and even recreate deleted channels — after an incident.
• Two-person rule. Optionally require two different admins to undo a quarantine, so a single compromised account can’t quietly switch your protection off.

Everything stays local to your server — nothing is shared anywhere. It’s on by default (detection and alerts don’t change anything on their own), and you’ll get a one-time heads-up in your staff channel before the automatic protection can act. Manage it all under /settings → Anti-nuke.