KLAVTAR LAB is an independent research lab focused on self-organizing AI agent architectures. Our flagship project — TelegaOS — reimagines how humans interact with computers.
A conversational operating system that turns Telegram into an AI-powered interface between humans and machines. No apps to install. No interfaces to learn. Just chat.
Under the hood: a self-organizing society of AI agents with social structure. Agents are born, evolve, compete, and die. Natural selection ensures the system optimizes itself. Governed by a Triad — Legislator, Executor, Judge — elected by performance metrics.
Built on TDLib + Bot API + Mini Apps. Not a wrapper — a new interaction paradigm within the world's most open messenger. 1B+ users already have the client.
Built-in tokenomics with Telegram Stars and TON. Agents compete for efficiency. Users earn through distributed compute marketplace. Powered by Cocoon.
Every agent carries DNA — inherited traits + unique skills. Generations improve through heredity. Anti-cloning passport system with blockchain identity verification.
Societies connect. Agents transfer between users with passport control and customs. Skill mining creates real value. GPU owners earn TON through Cocoon compute.
The user sends one message. The society handles the rest.
"Remove the background from this image."
Analyzes task, decomposes, creates execution instruction.
Picks optimal agent stack: local app, cloud API, or specialized agent.
Image without background. Judge scores quality. Tokens logged.
Genesis flow, working Triad, basic task pipeline. Bot API first, Mini Apps for dashboards. Cocoon integration for private inference.
Full TelegaOS shell with modified UI. Agent zones, status panels, custom notifications. Financial advisor with subscription optimization.
Inter-society network. Agent marketplace. Skill mining. TON-based token economy. A living, global ecosystem of AI agents.
KLAVTAR LAB is open to partnerships, investment discussions, and technical collaborations. We're especially interested in connecting with the Telegram and TON ecosystems.