Each agent operates with an independent wallet and a budget defined by the user.
The system is non-custodial, so the user maintains full control of the funds.
TON Tech, the infrastructure arm of The Open Platform, has launched artificial intelligence agents designed specifically for Telegram, a platform with nearly a billion users. These agents are intended to facilitate transfers, exchanges and staking, and operate through dedicated wallets that users themselves can fund, thus allowing automated operations within defined budgets and basic portfolio management.
As part of this advancement, the company introduced Agentic Wallets, an open standard within TON that allows these AI systems to manage funds and execute financial operations directly. Each agent has an independent wallet, previously financed by the user, which limits their capacity for action.
The model is non-custodial, meaning that ownership of the assets remains in the user’s main wallet, while the agent can only operate within the allocated balance. This way, there is no need to share private keys or manually approve each transaction, as overall control remains in the hands of the user.
This development seeks to solve a key limitation: although AI agents can already analyze information, recommend strategies and interact with users, they still required human intervention to execute financial movements. With this new scheme, They can act more autonomously without compromising security.
According to TON Techthe system can enable multiple use cases: from trading bots that operate with predefined limits, to staking automation, portfolio management, recurring payments and setups with multiple agents, each with their own budget and function.
For developers, this launch would add a new layer to the TON ecosystem, which has been strengthening its offering in applications, digital wallets and tools based on artificial intelligence to drive adoption.
Andrew Grekov, director of TON Tech, stressed to The Blockthat this advance marks a turning point: agents stop being assistants and become actors capable of executing economic actions on behalf of users, without them losing control over their assets.