Telegram community security · Private beta
Telegram anti-spam bot, built for AI spam.
An anti-spam bot that shows its work.
Most Telegram anti-spam bots are keyword filters with a settings page. Targa is an AI-powered screening engine, and every action carries a structured reason you can pull with /why <id>.
What it catches
Homoglyph obfuscation
"c0inbаse" reads fine to a human and slips past blocklists. Targa normalizes lookalike characters before scoring.
Edit-after-post
Clean message in, spam edit later — a classic filter bypass. Targa re-screens on edit.
Admin impersonation
Lookalike names and avatars targeting your members in DMs and replies.
Floods
Burst attacks are contained as a pattern, not fought message-by-message.
What else it catches
Reads what’s in the picture
Photos, stickers, image files and video covers are read like text. A pitch typed into a screenshot, or a link hidden in a QR code, is checked exactly like a plain message. English, Russian and Chinese, plus the language your group is set to.
Anti-raid join defense
Watches join velocity against your group's usual rate: when many accounts arrive within a short window, the newcomers are held muted for a moment and the admins are asked. A raid is flagged on evidence — ban-listed joiners, the same message from several of them — never on volume alone, and the copies are removed, the first ones included.
Conversational scam detection
Detection that reads the conversation, not just the single message: Targa sees the last few dozen messages and the sender's record, so a newcomer who chats first and pitches later is judged on the pitch in context.
Personalized phishing detection
Catches phishing written specifically for your group — right topic, right jargon — typically generated with AI. Targa judges intent with your group's own context: what is normal there, the sender's record and the conversation around the message, so a lure dressed in your community's language is still read as a lure.
Coordinated-campaign detection
Accounts that post the same pitch are linked: the same message from several new accounts in one group, or from several accounts across protected groups within an hour, is treated as one campaign, and the copies already posted are removed too. A join burst is held and shown to the admins with the evidence.
Any language
Targa reads meaning, not keywords, so it judges a scam in Portuguese, Hindi or Farsi as readily as in English.
How a Telegram anti-spam bot works
A bot is a member with admin rights. What it can do follows from that: it sees what an admin sees, and it can delete and ban. The difference between bots is what they do with what they see.
What it sees
As an admin, a bot receives every message in the group and every edit. It does not see members' private chats, and it cannot read a group it has not been added to. With the right to approve join requests it also sees who is asking to get in.
The door
Before an account can post, the bot can check it: a join CAPTCHA, a look at the account itself, and the pace of arrivals. Many accounts joining within a minute is a raid, and Targa judges the wave, not each account alone.
The room
Every message is read, including the picture, the QR code in it and the edit that arrives an hour later. Lookalike characters are normalized before judging, so "c0inbаse" is read as what it imitates. A friendly newcomer whose questions turn into an offer is judged as a conversation.
The action
A verdict becomes a delete, a mute or a ban, and every one carries a reason: /why <id> shows the verdict and the signals behind it. One tap undoes it. Targa enforces from the first message; an admin who wants to watch first sends /shadow on and reads the log instead.
Set it up in five minutes
- Request access. Targa is in private beta. Request access at app.targa.dev/register; the dashboard is where you connect a group and read its log.
- Add the bot to your group. Add @HeyTargaBot to the group like any member.
- Make it an admin with two rights. Delete Messages and Ban Users. Leave Add New Admins and Remain Anonymous off. Set its member tag to "bot" so nobody mistakes it for a human admin.
- Connect the group in the dashboard. The group appears in your dashboard once connected. From there you see every action, its reason, and the settings for join checks, floods and newcomers.
- Read the first reasons. Targa acts from the first message. Open /why on the first few actions to see what it judged and why; /undo reverses any of them. If you would rather watch for a week before it acts, send /shadow on.
What Telegram's built-in tools do, and where a bot starts
Telegram gives every admin a few settings. They are worth switching on, and they stop at the message itself.
| Built-in tool | What it does | What it does not judge |
|---|---|---|
| Slow mode | One message per member per interval. Takes the speed out of a flood. | What the message says. One scam per minute still lands. |
| Permissions | Switches links, media and stickers off for all members at once. | The account behind the message, or a scam typed as plain text. |
| Aggressive Anti-Spam | Telegram's own filter for groups over 200 members, with mistakes reported from Recent Actions. | Its rules are not published, and it gives no reason for a removal. |
| Member approval | An admin, or a bot, looks at an account before it gets in. | What the account posts a week later. |
A bot starts where these end: it reads the message, the picture and the edit, judges the account and the request, and tells you why.
What to look for in a Telegram anti-spam bot
Every message, every edit
Spam that arrives as an edit, or as the fifth message of a conversation, is the spam a first-message check misses.
Pictures and QR codes
A pitch typed into a screenshot, or a wallet link inside a QR code, contains no word a filter can match.
The account, not only the words
A lookalike name, an avatar copied from an admin, an account that joined this minute: these are the signals a word list cannot see.
A reason for every action
If you cannot read why a member was removed, you cannot tell a mistake from a catch. Ask for a reason you can show the member, and an undo.
The door as well as the room
Join requests, a CAPTCHA and a raid response keep the obvious accounts out, so the reader can spend its attention on the rest.
Your members' languages
A bot that matches English keywords misses the same scam in Portuguese, Hindi or Farsi. Targa reads meaning in any language.
Why explainability matters
Opaque moderation is the universal admin pain point: a member gets banned, nobody knows why, and trust erodes on both sides. Every Targa verdict is reproducible — verdict, score, and the exact signals that fired. Appeals stop being arguments.
Targa is AI-powered, so it also catches what no rule describes — conversational scam funnels, personalized phishing — and those verdicts carry the same /why.
Common questions
What is a Telegram anti-spam bot?
A bot you add to your group as an admin so it can remove spam and restrict the accounts that post it. Most are keyword filters with a settings page. Targa is AI-powered: it reads every message and every edit, and every action carries a reason you can read with /why.
Is there a free Telegram anti-spam bot?
Yes. Targa's free plan is genuinely free: not a trial, and no card required. It runs every detector. Paid rungs buy more channels, more message volume and more seats, never a different set of detectors.
How do I add an anti-spam bot to my Telegram group?
Request access at app.targa.dev/register (private beta), then add @HeyTargaBot to your group as an admin and connect the channel from your dashboard.
What admin permissions does a Telegram anti-spam bot need?
Two: Delete Messages and Ban Users. Everything else can stay off — including Add New Admins and Remain Anonymous, which should never be granted. Set its member tag to "bot" so the corner label doesn't read "admin". Added without admin rights, the bot stays a passive observer and still computes your Targa Score at member depth.
What kinds of spam does it catch?
Fake admins, lookalike characters, messages edited into spam, scams hidden in pictures and QR codes, link and crypto giveaway spam, floods and raids. And the attacks no rule describes: conversational scams that turn at message five, phishing written for your group, coordinated rings. In any language.
Will it conflict with my existing moderation bot?
No. Targa focuses on screening messages and members; rule-based moderation bots can run alongside it.
Does an anti-spam bot read private messages?
No. A bot sees the groups it has been added to. Private chats between members are invisible to it, which is why a scam that starts in the group and moves to private messages has to be stopped at the group: Targa flags the account that copies an admin's name or photo, and your pinned notice tells members that admins never message first.
What happens when the bot is wrong?
Every action carries a reason you can read with /why, and one tap on /undo reverses it. The member is back, and you can tell them what the bot saw. Targa enforces from the first message; /shadow on switches it to watching only, for an admin who wants to read the log first.