Updated: 2026-07-24
A moderator bot for a Telegram group keeps order on its own: it deletes messages with stop-words, catches flooding, escalates punishments from a warning to a mute, handles join requests and welcomes newcomers. In Corpis this takes no code: the same agent that replies in chat is added to the group as an admin, and you set the policy in the "Moderation" section. Against spam it uses a stop-word list and flood protection, and the irreversible ban is never placed by the bot itself — it goes into a queue for your confirmation.
Create or open an agent. Take an existing agent or create a new one: name, personality, rules and greeting. You can describe what you need in words and the AI assistant will fill the form.
Connect a Telegram bot. On the "Telegram" tab paste a bot token from @BotFather and click "Connect". One agent — one bot.
Add the bot to the group as an admin. Inside Telegram itself, add the bot to the group and give it admin rights — without this Telegram does not deliver messages to the bot and moderation does not work.
Open the "Moderation" section. In the agent settings go to "Moderation". If the bot is not connected yet, the platform prompts you to visit the "Telegram" tab first.
Turn on moderation and set the rules. Set stop-words and flood protection, punishments and the warning → mute → ban thresholds, the join policy and exceptions. Click "Save". You can sketch the rules in words — the AI returns a draft policy, but it cannot enable bans or an automatic mode.
Enable moderation for the group and watch the queue. In the per-group rights block, flip the moderation toggle for the group you need. Then watch the confirmation queue (bans and requests) and the action log, where mutes and bans can be reverted.
Order in the group is kept by a deterministic engine, not by the model: the target of a punishment is computed by code from the update, not by the AI. Only reversible actions run automatically — deleting a message, a time-boxed mute and a silent warning counter. The irreversible ban is NEVER placed by the bot itself: it goes into a queue, and you apply it with the "Confirm" button.
Protected targets are hard-blocked and fail-closed: the bot itself, users on the exceptions list and the group admins are never punished. While the bot has just become an admin and has not yet synced the admin roster, everyone is treated as protected — the engine does not ban blindly.
For the bot to moderate, add it to the group and give it admin rights — rights are granted inside Telegram, not in Corpis. As soon as the bot becomes an admin, Telegram turns off privacy mode and starts sending the bot every message so it can check them. In the "Moderation" section the real rights per group are shown read-only.
So the bot does not flood the group, in groups it replies only to a mention, a reply to its own post, or a command. This reply gate works at all times, even when moderation is off, while deleting and punishments work independently of it.
Bans and join requests wait in a confirmation queue — you confirm or reject each action. Applied actions appear in the log, and mutes and bans can be reverted (a deleted message cannot be restored). You turn moderation on for a specific group with its own toggle — you can bring order to one group and leave the others alone.
No. A ban always requires your confirmation: the engine puts it into a queue and you apply it with a button. Only reversible actions run automatically — deleting a message, a time-boxed mute and a warning.
You need an agent with a connected Telegram bot, added to the group with admin rights. After that turn moderation on in the "Moderation" section and, with a toggle, for the specific group. Rights are granted to the bot inside Telegram.
The bot itself, the group admins and members on the exceptions list are protected — destructive actions against them are hard-blocked. While the admin roster has not synced yet, everyone is temporarily protected: the engine does not punish blindly.
Yes. Stop-words, flood protection, punishments, thresholds and the join policy are all set in the "Moderation" section of the agent settings. You can describe the rules in words, and the AI assembles a draft policy for you to review and save.