🗺️ Roadmap

Where Server Assistant is heading. Priorities shift based on what server owners actually request — open a ticket via /support to suggest features or vote on what’s next.

📌 Dates are deliberately omitted. Things land when they're ready.

💚 Core moderation is free forever. AI features are metered with a 150K-token trial; Premium adds a larger monthly allowance plus extra capabilities. See pricing →

🧭 Three buckets, so you always know where things stand: 🟢 Coming up — committed, building now or next · 🔵 Future — planned, direction is set, timing isn't · 🔮 Potential ideas — exploratory, not yet committed. Your votes move things up.

🏷️ Version tags show release order, not dates. Each card carries the release it's targeted for — v5.6 ships before v5.7, and so on. We're currently on v5.5.4. Targets can shift as priorities change; Potential ideas stay unversioned until they're committed.

✅ ShippedLive in production today 🚀

Pulse ✅ Shipped v3.5↑0 Always-on AI server-health analyst. Daily digests, leading-indicator flags, channel-tone shifts. Never auto-actions — just gives staff x-ray vision. v1 shipped 2026-05-28 — 24-hour summary, week-over-week deltas, two leading-indicator flags (AutoMod-rate spike, sub-raid join spike). Channel-tone analysis lands in a later iteration.
Self-trained AutoMod ✅ Shipped v3.6↑0 Bot watches your staff's manual decisions (deletes + reported messages), clusters the patterns, then proposes server-specific AutoMod rules for owner approval. v1 shipped 2026-05-28 — opt-in observation, on-demand AI rule suggestions, accept/skip review. Nothing is ever auto-enforced.
Bot health insurance ✅ Shipped v3.7↑0 Bot watches its own impact in each server and proactively suggests tuning when it's becoming too noisy. Stops Server Assistant from quietly becoming the bot every server eventually mutes. Shipped 2026-05-29 — weekly self-check on AutoMod footprint, gentle tuning suggestion to staff chat, read-only, ≤once/14 days.
Threat Score ✅ v1 (advisory) shipped v3.8↑0 Risk-based moderation that replaces fixed punishment ladders. Actions fit the actual threat — context-aware, opt-in, with extensive safeguards against punishing legitimate users. v1 (advisory) shipped 2026-05-29 — free for every server — each user gets an explainable, time-decayed risk score shown to staff in their profile. It never auto-actions; it's pure visibility. The active, score-driven phase (Premium, with the safeguards below) comes only after the advisory version proves accurate in real servers.
Safeguards against false positives↑0
  • Sentiment + content paired with velocity (eager new members score down, not up)
  • First-week grace period for genuinely new users
  • Moderate scores trigger a friendly check-in DM, not an action
  • Score decay over sustained good behaviour
  • No silent kicks or bans — high-score events ping staff for review
  • Full transparency: every action shows its score breakdown to user + staff
  • Server-defined thresholds; fallback to fixed-ladder mode supported
🎫 Ticket Panels ✅ Shipped v5.2 Shipped 2026-06-03. In-server private support tickets — no DMs, no messy threads. Admins run /tickets setup to post a panel with a Create Ticket button; members open a private ticket-NNNN channel via a short intake form, and a full .txt transcript is saved on close. Free: one-question intake, private channels, transcripts. Premium: up to 4 custom questions, custom welcome/close messages, and adding users mid-ticket.
Baseline admin: /role · /lockdown · /tempban Baseline admin✅ Shipped v5.3 Shipped 2026-06-03. Three plug-and-play admin essentials — all permission-gated and written to your audit trail. /role add|remove changes a member's roles in one command, with guards so staff can't assign a role at or above their own. /lockdown start|end locks every channel at once for raid response and, on end, restores each channel's exact prior permissions — not a best guess. /tempban @user 7d bans for a set time and un-bans automatically, with a timer that survives restarts.
AutoMod AI second-opinion + repeat-offender detection ✅ Shipped v5.4 Shipped 2026-06-03. Two safety upgrades. AutoMod AI second-opinion (Premium, opt-in via /automod → AI Review) scores borderline filter hits and routes low-confidence catches to a one-tap staff review card — Delete & warn / Allow — instead of auto-deleting, cutting false positives. Repeat-offender detection (free, opt-in via /altguard) fingerprints banned/kicked users and, on a confident match, auto-bans likely ban-evading alts on join while alerting staff.
📩 Ban reasons & one-reply appeals ✅ Shipped v5.5 Shipped 2026-06-03. Every ban — manual, /tempban, AutoMod, or native — now DMs the user the reason and invites one reply to appeal. That reply opens a staff appeal ticket with Unban / Deny / Research buttons; Research runs Message Report (the renamed right-click AI summary) on the user's last message so you can judge context first. Soft-bans excluded; on by default.

🟢 Coming upNext — shaped by your votes

AI rule explainer 🎯 v5.6↑0 AutoMod DMs the user not just which filter matched, but why this rule exists, in plain language.
Smart purge 🎯 v5.6↑0 purge ai 50 deletes the genuinely problematic messages and archives constructive ones to a thread, instead of nuke-everything.
Native-action coaching Baseline admin🎯 v5.6 Logging of native Discord actions (right-click ban/kick/timeout) shipped in v5.1 — your audit trail is already complete. What's coming next: when a staff member moderates from Discord's own UI, they get a gentle one-line tip on doing the same thing through Server Assistant (so you keep your escalation ladder, approvals, and undo window intact). Read-only; never blocks the action.
🔵 Future Planned and prioritised — the direction is set, the timing isn't. Free-tier upgrades target the v5.6–v5.9 series; new Premium capabilities target v6.x. Voting reorders what lands first.
Mod team retrospectives 🎯 v5.7↑0 Weekly auto-generated retro embed: case counts, resolution times, notable patterns, items worth discussing.
Cross-server reputation 🎯 v5.8↑0 Opt-in positive reputation that follows trusted users between partner servers. The inverse of federated banlists.
Community-of-practice sharing 🎯 v5.8↑0 "Servers like yours configure X this way" recommendations based on similar community profiles.
Staff coverage map 🎯 v5.7↑0 Time-zone-aware visualisation of when each mod is typically active; flags coverage gaps for raid-prone hours.
Bot-on-bot detection 🎯 v5.9↑0 Flag suspicious automation pretending to be human — relevant for mass-marketing raids that evade text filters.
🟡 Premium-tier — new paid capabilities Part of Future. Core moderation stays free forever — the items below are new capabilities planned for the Premium tier, alongside the larger monthly AI-token allowance Premium already includes.

Tier A — ship first 🎯 v6.0

🌟 Premium Plus ($14/mo) Under considerationFor high-volume servers that push past the 750K/month Standard Premium cap. Includes: 2M tokens/month, priority Concierge response queue, multi-server bundle (3 servers per subscription), and bonus features as they ship. Pending validation from 50+ Standard Premium subs — highest-voted ideas from that cohort will define the final feature set.
Reaction roles↑0Button-based self-service role panels in any channel.
Custom slash commands↑0Server owners define their own /foo returning a configured embed or text snippet.
Daily AI channel summaries↑0Once-a-day digest of activity in a chosen channel, posted to a configured destination.
Mediator↑0When a conversation is escalating before any moderation action, bot DMs both parties, collects each side privately, and produces a neutral summary for staff. (Distinct from one-reply ban appeals, which already shipped in v5.5 — this is the pre-action sibling.)
Decision explainer + multi-round appeals↑0Builds on the one-reply ban-appeal flow shipped in v5.5. Adds: warned/muted users can DM the bot "why was I actioned?" for a plain-language answer drawn from the staff-supplied reason and the matched rule; banned users get a guided multi-round appeal (a short AI-facilitated back-and-forth that gives staff more than a single line to judge).

Tier B — solid additions 🎯 v6.1

Leveling / XP↑0Server-wide XP with level-up roles and a leaderboard. Opt-in per server, opt-out per user.
Polls↑0/poll with button-based voting and live tally.
Embed builder↑0Interactive wizard for crafting rich announcements with fields, images, and buttons.

Tier C — nice-to-have 🎯 v6.2

Birthday tracker↑0Opt-in birthdays with auto-DM and channel post on the day.
Welcome card images↑0Auto-generated PNG welcome for each new joiner.
AI Q&A from server FAQ↑0Bot answers questions based on FAQ snippets the owner has provided.
Auto-translate↑0On-demand or automatic translation for configured channels.

Tier D — power users 🎯 v6.3

REST / webhook API↑0Read-only endpoints for your guild's bot data (warnings, audit log, activity stats).
White-label branding↑0Fully replace the bot's identity per-server — your name, your avatar, your colour scheme.
Unlimited /imagine↑0Premium removes the 30-second per-guild rate limit.
🔮 Potential ideas Bigger conceptual swings we're genuinely interested in — including voice-channel moderation, the industry's biggest blind spot. Exploratory, not committed to any release. The ones you vote up are the ones we investigate first.
Voice-channel moderation↑0Real-time transcription + AutoMod / threat-score logic applied to voice channels. The industry's biggest moderation blind spot — no major bot touches voice.
Federated moderation network↑0Opt-in cross-server ban / warn data sharing for trusted communities. Solves cross-server raid coordination.
AI-drafted server policy↑0Bot reads your channels for two weeks, drafts a rules document tailored to your community's actual norms, posts it for owner approval.
Conversational onboarding↑0Replace the static welcome DM with a short AI conversation that auto-assigns roles based on answers and filters bad-faith joiners before their first message.
Drift detection↑0Track server tone and behaviour over weeks. Alert the Owner when material drift is detected so culture can be maintained before it erodes invisibly.
Community DNA fingerprint↑0Per-server embedding capturing community character. Powers new-member matching, cross-server compatibility, and rule recommendations.
Predictive mod scheduling↑0Operational intelligence: "Sunday evenings see 3× your average AutoMod hits — adjust on-call rotation?"
Mod-team coaching / bias audit↑0Compares each mod's decision patterns against the team average; surfaces outliers for self-reflection, never public shaming.
Persona / role-based moderation↑0Different filter aggressiveness per role tier. Community norms scale with trust — verified members get more latitude, unverified get less.
Channel-specific tone tuning↑0Per-channel policy — vent channels soft, family-friendly strict, debate channels allow heat but flag personal attacks.
Misinformation flagging↑0Fact-checkable claims get a 💡 reaction linking to a verified source. Never auto-removes; treats the user as an adult.
Reverse moderation↑0When a conversation derails, the bot suggests a constructive question or topic shift — soft-touch alternative to message deletion.
Cross-modal moderation↑0Unified text + image NSFW + link safety + voice analysis. Closes the gap where bad actors flip between modalities to evade single-mode filters.
AI-generated mod training↑0Auto-generated new-mod onboarding doc based on actual cases your team has handled and house style they've established.
AI thread management↑0Off-topic conversations get auto-threaded with a soft DM to participants — "moved this fascinating tangent to its own space."

📣 How to influence the roadmap

The fastest way to move something up the list:

  1. Vote on the cards above — the highest-voted ideas get prioritised.
  2. Open a /support ticket describing the use case — not just “add feature X” but what you’d do with it.

What ships is what gets requested most clearly. Vague “add more features” feedback is unactionable; “I run a 2,000-member RP server and we need X because Y” gets prioritised.

What’s next?