Pre-launch

Telegram captcha

Stop it at the door, and keep the receipt.

A captcha is the cheapest filter you own. It is also the one most likely to quietly turn away a real member, and the one least likely to tell you it did.

The part that goes wrong

A door that never reports back.

Most captcha bots are a gate and a counter. They tell you how many accounts were blocked. They do not tell you which real people bounced off the step and never came back, which is the cost you actually care about and the one nobody measures.

A join filter that cannot be reviewed is a growth problem wearing a security badge.

Why do Telegram captcha bots turn away real members?

Because a captcha is a blunt instrument aimed at behaviour, not at people. It asks everyone the same question at the worst possible moment, when a member has just arrived and has no reason yet to work for entry. Some of them close the app instead, and a bot that only counts blocks has no way to show you that.

The failure is not that the filter is imperfect. Every filter is imperfect. The failure is that it is imperfect and silent, so the one number you can see is the flattering one. Blocks look like work being done. The people who gave up look like nothing at all, because nothing is what was written down about them.

What is different here

The door is part of the record.

A captcha decision is an action, so it carries what every action carries: the community, what detected it, the evidence, the action taken and who owned it. Appeals and undo are retained, and the reversal is recorded too.

It also sits under the same policy as the rest of your moderation, rather than in a separate bot with its own settings page and its own idea of who is allowed in.

What does SwarmUp record when the captcha stops someone?

The same fields every other moderation action carries, because entry control is not treated as a special case:

  • the community
  • the sender
  • what detected it
  • a confidence value
  • the evidence
  • the action taken
  • who owned it

That list is the difference between a counter and a record. A counter answers how many. A record answers which one, on what evidence, decided by whom, and whether anybody changed their mind afterwards. Only the second kind survives being asked about weeks later by a member who says they were shut out for no reason.

Raids do not stay on one platform

One rulebook, both platforms.

A raid that starts in your Telegram group is usually already queued against your Discord. Two bots on two platforms with no shared record means the second one meets it as a stranger. SwarmUp runs both from one workspace, so the door learns in both places at once.

Can one captcha policy cover both Telegram and Discord?

Yes. Both communities are operated from the same workspace, and the rules they run on come from one Community Blueprint rather than from two unrelated settings pages. Each platform and each channel still keeps its own permissions, so shared policy does not mean identical behaviour everywhere. It means one place to change your mind.

The practical version of that: when you tighten the door because of something you saw in Telegram, you are not then remembering to go and do the same thing in Discord before the weekend. The half of an incident response that gets forgotten is always the half that lives in the other tool.

What it costs to keep the door shut

Protection is not the upsell.

The usual shape of this market is that the door is free and everything that lets you understand the door is not: history behind a tier, appeals behind a tier, a second group behind a per-server price.

Does a Telegram captcha bot cost more as the group grows?

Not here. Community growth is not metered, core protection is not a premium feature, and only AI generation you ask for uses credits. Servers, groups and seats are unlimited on every plan, including the free one, so the bill does not move because more people arrived or because you put a door on a second group.

This matters most on exactly the day it is hardest to think about it. A raid is a spike in volume, and a bill that follows volume charges you for being attacked. The plans are priced on how many workspaces you run, which is a number you choose, rather than on how much trouble showed up at your door, which is not.

Where the automation stops

What the AI is not allowed to do.

The reason to trust a moderation product is not how clever it is. It is what it is prevented from doing without you.

  • AI cannot ban a member on its own.
  • AI cannot apply a permanent sanction on its own.
  • AI cannot publish a change to your policy.
  • Deterministic low-risk rules act automatically. Everything above that line waits for a person.
  • Every action is reversible, and the reversal is recorded too.
  • Provider identities are stored as one-way pseudonyms, not as raw platform accounts.

Straight answers

The questions that come first.

What is a Telegram captcha bot?

A Telegram captcha bot puts a challenge in front of anyone joining a group, so an account has to complete a step before it can post. It filters automated joins at the door. SwarmUp runs one as part of your moderation rather than as a separate bot with its own settings page.

Does a captcha stop a raid on its own?

No. A captcha filters automated joins, and a coordinated raid using real or aged accounts can walk straight through one. It is the cheapest filter you own, not the whole defence, which is why the door here runs under the same policy and the same incident queue as the rest of your moderation.

Can a member appeal being blocked at the door?

Yes. A captcha decision is an action, so it carries an appeal and an undo like every other action. The reversal is recorded too, which means you can see both that somebody was stopped and that you later let them in, rather than only the first half.

Can the AI ban somebody for failing the captcha?

No. AI cannot ban a member on its own and cannot apply a permanent sanction on its own. Deterministic low-risk rules act automatically, and everything above that line waits for a person. Failing a challenge at the door does not turn into a permanent decision without somebody making it.

Does SwarmUp store the Telegram accounts it checks?

Provider identities are stored as one-way pseudonyms rather than as raw platform accounts. The record keeps what detected something, the evidence, the action taken and who owned it, so a decision stays reviewable later without the workspace holding a copy of your members' platform identities.

Can I put a captcha on more than one Telegram group?

Yes. Servers, groups and seats are unlimited on every plan, including the free one. What separates the plans is how many workspaces you run, not how many communities or members you have, so putting a door on a second group does not become a line on the bill.

Where we are today

SwarmUp is not open to the public yet. Moderation is the core of the product and it runs today. Engagement and rewards are still in testing. Plans and prices are final, and nothing is charged before the workspace opens.

One email, when it opens

We will not waste the one email.

We are not naming a launch date until we can hold it. What you get for an address is a single message the day the workspace opens, and the same plain account of where the product stands on that day.