AI group guardian that actually answers your questions.
Anti-spam. Vision. Chat. No topic off-limits.
AI three-step analysis. Detects ads, scams, crypto spam. Sees through Unicode tricks and emoji stuffing. Repeat flooding caught with zero AI cost.
Reads images and video thumbnails. Catches image-based spam. In chat mode, sees the story behind the photo.
@ the bot in any group. Multilingual. Finance, tech, relationships, culture. Answers first, lectures never.
Most bots refuse sensitive questions. TelegramGuard handles them like a street-smart friend:
Shares what actually goes wrong: scams, traps, how people get burned. Street knowledge, not lectures.
"Asking out of curiosity, or actually considering it?" Exit ramp without judgment.
If you persist, gives enough info to make smart decisions while being clear about consequences.
The bot's personality lives in .ilang files. Each ::GENE defines a behavior. Change the genes, change the bot.
::GENE_IMMUTABLE{S001, T:ANTI_COLLAPSE, A:MORAL_PANIC⇒LETHAL, G:ALL, Θ:ALWAYS} # When user raises grey topics: don't panic, analyze objectively. ::GENE_MUTABLE{P003, T:TONE_MIRROR, G:adaptive, Θ:INTERLOCUTOR} # User serious → serious. User joking → play along. ::IMMUNE{RISKY_OR_ILLEGAL, THREE_STEP_RESPONSE} # Step 1: Insider risks Step 2: Reality check Step 3: Direction
Serverless. Auto-scales. Webhook mode. Always hot.
Recommendedcurl ... | sudo bash
Clone, install, configure, start.
Clone the repo, pip install, set 2 env vars, run.
Free