Always on
The bot is awake every minute of every day. New spammers get caught the moment they post, not the next morning when an admin scrolls up.
Case study · Always-on community bot
A custom Telegram bot that handles everything a busy private community needs — moderation, security, financial approvals, growth reporting — so the human admins only get notified when a human is actually needed.
The problem
A growing private Telegram community had outgrown its admin team. Spammers were constantly trying to break in. Crypto scammers impersonated members. Routine messages flooded the chat. The admin team was tied up rejecting fake withdrawal requests, approving real ones, banning scammers one at a time, and trying to keep the conversation civil at 3am. The community needed help with the routine work so the humans could focus on the conversations that actually mattered.
The approach
Rather than gluing together five off-the-shelf moderation tools, the right move was a single custom bot designed for this community's actual rules. Every behaviour was decided by the admins, not assumed by a template. The bot was built to be quiet by default — only speaking up when there was something a human needed to know — and to keep a full record of every decision it made, so nothing it did could be lost or disputed.
Built right
A community moderator who never sleeps, never gets bored, never bans someone in anger, and keeps an audit trail of every single action it takes.
The bot is awake every minute of every day. New spammers get caught the moment they post, not the next morning when an admin scrolls up.
Every action — every ban, every approval, every notification — is logged with a timestamp. Admins can look back at anything that happened, weeks later, and see exactly what was done and why.
The bot will never mute or ban another admin's message. The high-stakes operations (financial approvals especially) only fire for people on an explicit allowlist, even if their admin status is technically sufficient.
If something unexpected happens, the bot stops and asks rather than guessing. Better a notification than a silent mistake.
What it does
Recognises and removes the patterns that flood community chats — bulk messages, link spam, “DM me to invest” outreach. Spammers vanish before the rest of the room sees them.
Catches accounts pretending to be admins or trying to lure members into private chats. Banned on sight, with a notification to the admin team explaining why.
New joiners prove they're human before they can post. Most bot accounts give up at this step and never make it into the room.
When a member requests a withdrawal of their in-community currency, the bot routes it to the allowed approvers, captures their decision, and either pays out or refunds — all logged.
Monitors the community's public currency market and posts a chat notification each time a notable buy happens. Adds activity to the room without anyone having to refresh a chart.
Admins can ask the bot for daily, weekly or all-time community-growth charts. Real numbers, on demand, no spreadsheet.
Outcome
The admin team's day-to-day work has shifted from “keep the lights on” to “build the things only humans can build” — community events, content, partnerships, fundraising. Routine moderation no longer competes for their attention. Financial approvals that used to involve coordinating across time zones now take one tap. The community has grown steadily, and the bot has scaled with it without needing anything rebuilt.
What I'd do differently
The very first version of Thorak just did its job silently. When admins later wanted to look back at why a particular ban had happened, the answer was “we don't know”. The audit log was added as a follow-up and instantly became the most-used admin feature. Lesson for any system that takes irreversible actions on behalf of a human team: build the receipt first, the action second.
Want a Thorak for your community?
Off-the-shelf moderation tools handle the average community. Custom is for the specific. Quoted fixed-price after a free 15-minute discovery call.
More case studies