Everyone’s Looking For The Exit
Every time Discord shows up in a headline, someone starts a thread asking where the community should move next. Half the replies argue about age verification. The other half argue about which self-hosted app deserves the crown. I looked at a lot of those threads, and read the Compose files behind the two self-hosted names that keep coming up, so here is the short version.
Most people asking “should we leave Discord” do not actually need to. They are reacting to a headline, not a problem their server has. A smaller group has a real reason: a raid team tired of being the product, a niche community that got flagged for nothing, a forum owner who wants their data on their own disk. For that group, the honest answer is blunt. Every one of you will lose members no matter which app you pick. Migration cost is a people problem, not a software problem. Pick the tool that matches how your community actually behaves, tell people plainly and early, and accept that a chunk of your server will not follow you anywhere.
That is the argument. Everything below is detail.
The Age Check Everyone Got Wrong
Discord announced global “teen by default” account settings on 9 February 2026, with a rollout meant to start in early March. That same press release got a quiet update on 24 February 2026 adding a delay notice, after a wave of backlash in the two weeks between. The stated reason, in Discord’s own words, was to “expand verification options, increase vendor transparency, and publish detailed technical documentation, while continuing to meet regulatory requirements where needed.” The global rollout moved to the second half of 2026, and as of September 2026 it still has not fully landed. Plenty of the panic posts you have seen describe a policy that has not shipped everywhere yet.
Here is Discord’s own claim, quoted directly rather than paraphrased, because the exact wording matters: “Over 90% of users will continue using Discord exactly as they do today, with their age automatically determined through our existing safety systems and no additional action required. We do not require universal ID uploads or face scans and only a small percentage of users, generally those seeking access to age-restricted spaces, may be asked to confirm their age.” For most accounts, an in-house model infers an age group from behavior and account signals, and Discord states plainly that it does not read message content to do it. A manual check only triggers on specific actions such as unblurring sensitive media, switching the content filter to “Show,” turning off message requests, or entering an age-restricted server. The UK and Australia already had age checks in place before any of this, for reasons that have nothing to do with Discord’s own policy.
When a manual check does trigger, the vendor is k-ID. One path is a video selfie for facial age estimation, processed on the device and never uploaded. The other is an ID scan handled by k-ID with document partner Veratad, which Discord says it never sees directly and deletes once verification finishes. Two more options, a Google Wallet passport pass and a small refundable credit card charge under 50 cents through Stripe, are in testing as of September 2026 and not available to everyone yet. That credit card option alone undercuts the “they want your face” version of this story that keeps getting repeated.
None of that makes the distrust irrational. Discord disclosed in October 2025 that a third-party vendor handling age-related appeals, not Discord’s own systems and not the main verification flow, may have exposed sensitive data for about 70,000 users, including government ID photos. That is a real breach with a real number attached to it, and it is exactly why people who read the narrow, careful policy above still do not want to hand any vendor an ID photo. Both things are true at once: the policy Discord actually shipped is far narrower than the panic version, and the appeals-vendor breach is a legitimate reason to stay wary of the pipeline regardless. Get both halves right before you write your community announcement, because the people who left over a garbled version of this story are going to ask pointed questions.
The Part Nobody Wants To Hear
Here is the uncomfortable math. A migration announcement gets your loudest members to show up on day one, out of curiosity more than commitment. Your lurkers, the people who read every day and post twice a year, are the ones who quietly stop showing up. They were never going to click a second invite link for a community they only half-engage with, and no replacement app changes that. Your real decision is how many you lose and which ones, because losing some of them is already settled.
Stoat: Discord With The Serial Numbers Filed Off
Stoat is the closest thing to a drop-in replacement. It was called Revolt until the project renamed itself in October 2025 after a cease-and-desist, moving from the revoltchat organization to stoatchat. The rename runs skin deep: the config file you mount into the containers is still named Revolt.toml, which tells you the team spent its time on the product rather than a cosmetic rebrand. It is licensed AGPL-3.0, the codebase dates back to 2021, and as of September 2026 the project sits at 3,281 stars on GitHub.
The channel list, the server sidebar, the emoji reactions, all of it reads like Discord because it was built to. That familiarity is why people pick it first: your moderators do not need a manual, and members who show up expecting Discord will not feel lost. Stoat also runs a hosted instance, so a community can try it with an account instead of a server first.
Pick Stoat if your community wants the Discord experience with a different name on the tin, and your moderators do not have the bandwidth to learn a new mental model on top of a migration.
Fluxer: The Newer, Sharper Option With A Catch
Fluxer is the more polished newcomer, also AGPL-3.0, built by a Swedish team. The repository was created on 1 January 2026 and by September 2026 it had reached 10,196 stars, roughly triple Stoat’s count despite being months old instead of years old. That kind of early traction says the project scratched an itch people had been waiting on.
The catch is real and worth saying plainly: Fluxer has no native mobile apps and no federation, as of September 2026. Both sit on the project’s roadmap with no ship date attached. If half your members live on their phones, which describes most communities, that gap is disqualifying today. Come back to Fluxer once mobile apps land, not before, if your server depends on people checking in from the couch or the bus.
Pick Fluxer if your community is desktop-first and you want the more modern interface, and you can live without mobile apps until they arrive.
A Word On Hosting It Yourself
Both Stoat and Fluxer are self-hostable, and both want more than a spare Raspberry Pi to run comfortably. Fluxer’s own Compose file declares memory limits that add up to about 16.75 GB, with roughly 5.38 GB reserved right out of the gate. Those are the caps the project asks Docker for, not numbers measured off a running deployment, and they tell you Fluxer wants a real server, not a hobby box in a closet. I read both Compose files line by line, service by service, and every number is in Stoat vs Fluxer: Self-Hosting Costs, including the one difference that actually matters if you plan to open a firewall for voice.
Matrix: The Federated Long Game
Matrix, usually deployed as Synapse with an Element client, is the protocol-first choice. Nobody owns your account, your room can bridge to other networks, and your history is yours to keep as long as you keep running a homeserver. That is the appeal for anyone who has been burned by a platform changing the rules on them before.
It is also the option that feels least like Discord. Rooms are not servers, spaces are not quite categories, and new members spend real time figuring out where things live. Voice and video are the weak spot. They work, and they will carry a meeting, but they trail what Discord trained your members to expect. People pick Matrix because they refuse to be locked to one operator again, and they accept the call quality as the price of that.
Pick Matrix if federation and not being tied to a single company matter more to your community than a smooth first ten minutes.
Mumble: The Honest Outlier
Mumble does not pretend to replace Discord’s text features, and that restraint is the point. If your community is really a voice channel with a chat log stapled to it, a gaming group that talks during raids and barely posts otherwise, Mumble is a better fit than any of the three chat-first options above. It is old in the way that matters: the protocol has had years to get low-latency voice right, and it runs on hardware so modest that “spare machine in the closet” covers it without exaggeration. Text channels, reactions, and community features exist but stay deliberately thin, licensed under a permissive BSD license rather than AGPL.
Pick Mumble if voice is the actual product your community uses, and the text channel is a place to post the raid time and nothing more.
The Four, Side By Side
| Stoat | Fluxer | Matrix | Mumble | |
|---|---|---|---|---|
| Feels like Discord | Closest match | Close, more modern | Least similar | Not trying to |
| Voice quality | Good | Good | Weakest of the four | Its whole reason to exist |
| Mobile apps | Yes | None yet (roadmap) | Yes, via Element | Official iOS, Mumla on Android |
| Federation | No | No (roadmap) | Yes, by design | No |
| Self-host difficulty | Moderate | Moderate to heavy | Heavy | Light |
| Licence | AGPL-3.0 | AGPL-3.0 | AGPL-3.0 (Synapse) | BSD |
If You’re Actually Doing This
Run both servers in parallel for a few weeks before you touch anything. Move your loudest channel first, not your quietest one, because the loud channel is where people notice the new place is alive. Name a specific cutover date and repeat it more than once; a vague “we’re moving eventually” trains people to ignore it. Expect to lose your lurkers no matter how well you communicate, because they were never going to click a second link. Keep the old Discord server standing as a signpost pointing to the new one instead of deleting it. A dead end with no forwarding address is how you lose the stragglers who show up three months late looking for their friends.
Just Stay On Discord If…
…your community’s only real complaint traces back to a headline. If nobody in your server has hit a moderation problem, a data ownership concern, or an actual outage that cost you something, the age verification news is a thin reason to spend a month migrating people who are perfectly happy where they are. Moving a community is expensive in attention and goodwill. Save that budget for a problem your members can actually feel, because this policy has not fully rolled out yet and mostly will not touch them when it does.
Common Questions
Is Discord requiring ID from everyone in 2026?
No. Discord says more than 90% of users will never be asked to verify anything, because age gets inferred from existing account signals. The global rollout announced in February 2026 was delayed to the second half of 2026 and, as of September 2026, still has not fully landed everywhere. Verification only triggers for accounts trying to reach age-restricted spaces.
Which Discord alternative is closest to Discord?
Stoat is the closest match. It was built as Discord’s near clone under the name Revolt before renaming to Stoat in October 2025, and the channel list, server sidebar, and emoji reactions are deliberately familiar. That familiarity keeps moderator training and member confusion to a minimum during a move, which is the main reason communities pick it first.
Does Fluxer have a mobile app?
No, not yet. Fluxer has no native iOS or Android app as of September 2026, though one is on the project’s public roadmap. For a community where members check in from their phones, that gap matters today. Communities that are mostly desktop-based feel it far less than ones built around phone-first casual chatting.
Will my community actually follow me off Discord?
Partly. Your active, engaged members generally follow within the first few weeks if you communicate clearly and pick a real cutover date. Lurkers, the people who read daily but rarely post, mostly do not click a second invite link for a server they only half-engage with. Expect a real drop in headcount regardless of which replacement you choose.
